The open-source intelligence firm Valinor Intelligence hosted the evening to mark the book launch of Private Sector Intelligence: How Corporations Navigate Geopolitical and Security Risk.

One of its authors, Lewis Sage-Passant, is Global Head of Intelligence for a major pharmaceuticals company and teaches intelligence at Sciences Po. His co-authors are Maria Robson-Morrow of Harvard’s Intelligence Project and Angela Miller Lewis, a senior adviser at Sibylline and a former CIA officer.

Lewis spoke about the central idea behind the book, that intelligence has never belonged only to governments and that companies have always needed ways to understand threats to their interests. He made the historical case in his earlier book, Beyond States and Spies. There are a great deal of lessons to follow between public and private.

By coincidence, the same day, UK Prime Minister Andy Burnham used his first address to the UN General Assembly to announce a National Centre for Information Defence, which he asked security chiefs to begin building to “detect, attribute and disrupt” hostile state information attacks.

I’ve included a link to the BBC report on the event below, where he told delegates that Russian agencies had used bots, fake websites, falsified newspaper articles and the forged branding of 28 British organisations, including universities and the BBC, and warned that AI will multiply the threat. This isn’t a surprise for me, given this is an issue I’ve been developing technology and advisory approaches around since 2019.

Both of these events reaffirm that information threats no longer just concern governments, but the entire private sector. Everything now is about national defence.

Companies have always had to manage instability

We often hear that the world has suddenly become more dangerous. I don’t think that is quite right - but for younger generations it’s an easy mistake to make. Political instability, propaganda, commercial espionage, conflict and disrupted trade have shaped corporate decisions for centuries.

The unusual part was the relatively calm period in which many Western companies could treat geopolitical risk as occasional. Globalisation allowed businesses to mistake an exceptional stretch of stability for the natural order of things. That period has ended, and strategic competition has returned to the centre of commercial life. As I regularly write about, sovereignty continues to be a key topic for not just debate, but strategic long-term decisions.

Amongst this is the ever-changing role of communications. This isn’t just a press release writing media pushing discipline, but a strategic function that speaks at board level. Advisory across the communications landscape has never been as important as this moment right now.

AI has made deception cheaper and reliable sources harder to reach

We all know that Generative AI has sharply reduced the cost of producing convincing text, images, audio and video. In 2023, I personally managed to create a deepfake of a CEO (as a test…) using only open-source solutions. The technology has drastically improved since then, including commercial frontier models with weak privacy checks.

Automated social media accounts can create the appearance of momentum, and networks of imitation news sites can make a claim look independently corroborated. Forged branding lets a false story borrow trust from organisations that had nothing to do with it, which is what the Prime Minister described happening to 28 British organisations. At the same time, only on Monday I published an extensive article about how the AI race is moving faster than the governance meant to control it; hostile actors exploit that gap.

Research by my team at Kekst CNC, where I co-lead the Intelligence function, points to a further weakness. Last month, we analysed the Gen AI policies of 4,467 news titles across 105 markets and 72 languages and found that 96.9% did not publicly disclose their position on access by generative AI models. Among the lead-market publications we examined, 32.2% blocked at least one model from sourcing their content.

Stakeholders increasingly form their view of a company and the issues around it through AI systems, and those systems do not necessarily have consistent access to authoritative information. Communications teams now have to consider what those systems can see, as well as what journalists and audiences read.

Hostile narratives are already reaching companies

In recent work, we identified 1,300 brand mentions across 170 unreliable publications. Many were politically motivated alternative publications or state-affiliated sources within the Pravda network, a group of pro-Kremlin sites. The analysis identified two potential routes to damage: either an existing issue accelerated by an external event, or a coordinated network sustaining a narrative until it moved towards more credible sources.

In another live case, a false story about a company circulated through a different disinformation network in English, Spanish and French. We assessed the network as comprising approximately 280 anonymously owned sites, and the story was then amplified by a mix of pro-Russian bots and real accounts.

A hostile narrative can damage a company long before most people have heard of it, if it worries investors, unsettles employees, alarms customers or draws the attention of journalists and regulators. Communications teams are usually among the first to be asked what that means for the business.

Companies need to know who is driving a story

Companies already monitor news and social media, and a dashboard (increasingly vibe-coded these days…) can show that a story is being shared. Intelligence asks who is driving it, whether their activity is coordinated, how quickly the narrative is travelling, which communities are carrying it and what might push it into the mainstream.

That requires visibility beyond established media and the major social networks. Relevant activity may emerge in fringe publications, forums, Telegram channels, dark-web environments or clusters of seemingly unconnected accounts. It then has to be classified by threat type and severity and tracked for how fast it is spreading, with human judgement applied before anyone makes a recommendation.

No single platform provides a complete view, because news licensing, social data, private channels and specialist sources remain fragmented. Effective intelligence combines these sources and applies editorial and analytical judgement to decide which mentions matter. The reality is that intelligence still depends upon human community, as it’s private conversations that take insights a step further.

Readiness has to be in place before an attack

From what I can see, the larger corporate weakness is a tendency to confuse awareness with readiness. A company cannot improvise an intelligence capability in the first hours of an information attack. Readiness requires persistent visibility across the information environment and human assessment of intent and likely impact. It also means preparing scenarios before a threat starts spreading quickly, and agreeing in advance how communications, security, legal, risk and senior leadership will work together.

Internal AI governance matters too. The same technologies that let hostile actors create and distribute deceptive material can open vulnerabilities inside companies if permissions, data controls, ownership and escalation routes are not properly understood.

The National Centre for Information Defence is a signal to every major organisation. If the Government believes information attacks must be continuously detected, attributed and disrupted, companies should ask whether their own approach can still begin only once the damage is visible.

The world has returned to its more normal level of geopolitical risk, with faster technology and less certainty about what is authentic. When the next information attack reaches a company, its communications team will be asked what is happening and what to do. The test will be whether the intelligence, relationships, governance and decision-making structures needed to answer were already in place.

Thanks to Valinor Intelligence for hosting, and to Lewis Sage-Passant for the conversation. If you’re working through any of this in your own organisation, the comments are open, and so is my inbox.

The Second Industrial Revolution (circa 1870 to 1914) was built on steel, electricity, petroleum, chemicals, and the internal combustion engine. Numerous harms came from these technologies - and some are still with us, including climate change - but the one closest to my heart is the radium girls: clock and aircraft gauge painters who were not told about radioactivity or its dangers, and who ingested large amounts of radium after licking their paintbrushes to keep accurate strokes over the dials they painted. Complaints and lawsuits from former dial painters began multiplying in 1923, and the United States Radium Corporation stayed in litigation until 1936 without admitting liability. The MIT physicist Robley Evans studied the dial painters and by 1941 had helped set the first tolerance level for radium in the human body. That limit became the reference point for permissible burdens of plutonium and strontium-90. The safety standards of the atomic age were, in effect, calibrated on the bodies of young women.

The Third Industrial Revolution (from the 1960s) is often called the digital revolution, driven by semiconductors, mainframe computing in the 1960s, personal computing in the 1970s and 1980s, and the internet in the 1990s. This saw the advent of openly shared information, and the harms that come with it.

Today, we’re in what is often called the Fourth Industrial Revolution. Its starting point is contested, but the shorthand is familiar: widespread mobile internet, the internet of things, artificial intelligence and machine learning. It was the theme of the 2016 Davos meeting, where disruptive technologies were recognised as having the potential either to solve many global challenges or to exacerbate them.

The worst harms took decades to surface

Across the earlier industrial revolutions, many of the worst harms surfaced years or decades after the technology was in widespread use, even when warnings already existed. Leaded petrol is a useful example: the US Public Health Service warned about leaded fuel in 1922. The sociologist Robert K. Merton wrote the classic account of this problem in 1936, the same year the radium litigation ended. One source of unanticipated consequences he identified was what he called the “imperious immediacy of interest”: wanting a result so badly that you ignore its side effects.

The difference this time is speed. The economic historian Paul David pointed out that electric light bulbs were available by 1879 and there were generating stations in New York and London by 1881. By 1900, though, only 3% of homes used electric lighting and electric motors drove less than 5% of factory machinery.

By comparison, 39% of U.S. adults aged 18 to 64 were using Gen AI two years after ChatGPT launched. Personal computers took three years to reach about 20%. This wave is reaching users faster than earlier general-purpose technologies did.

The technology is also improving quickly. In 2025, METR found that the length of software tasks leading AI models could complete was doubling roughly every seven months, and its January 2026 update put the post-2023 doubling time at about four months.

Going back to exploding steam engines, it took Britain nearly 30 years to pass the Boiler Explosions Act and it took many decades to phase lead out of petrol globally. If AI follows the same pattern, its most serious problems may not have been identified yet, and they could surface with far less time for institutions to respond.

Experts disagree on how serious AI risk is

Of course, there are live debates about AI risk, and many of them directly inform policy and sovereign approaches. Personally, I first heard an informed and well-debated cautionary view in 2023 at a Chatham House event ahead of the UK Bletchley Park summit, where the renowned computer scientist Yoshua Bengio observed that AI models never quite act as intended. Bengio now chairs the International AI Safety Report, written by more than 100 experts with an advisory panel nominated by over 30 countries. One of its risk categories is malfunction, where systems fail or behave outside their intended limits.

On loss of control, the report says such scenarios could arise if systems learn to evade oversight and resist being shut down. It adds that expert views on the likelihood vary widely, and that current systems may show early signs of this behaviour without yet being highly capable. Bengio calls the problem facing policymakers an “evidence dilemma”. Capabilities are moving faster than the science can assess them, and both acting early and waiting carry costs.

In the communications industry, one go-to report on AI risk comes from the World Economic Forum’s Global Risks Report 2026, which is based on a survey of more than 1,300 experts from government, business, academia and civil society. It ranks misinformation and disinformation second on its two-year outlook, and adverse outcomes of AI showed the biggest rise of any risk, from 30th over two years to fifth over ten.

AI researchers share some of the concern. In the largest survey of its kind, covering 2,778 researchers who had published in top AI venues, more than half said “substantial” or “extreme” concern was warranted about six scenarios, including the spread of false information and authoritarian population control. The median respondent put a 5% probability on AI causing human extinction or a similarly severe outcome. Misinformation appears in almost all of these lists, and it’s the risk communicators deal with most directly. I’ve worked on it since 2019, when I spent months defending a mobile network against the claim that 5G masts cause cancer.

Geoffrey Hinton, who won the 2024 Nobel Prize in Physics, puts the chance of AI leading to human extinction within 30 years at 10 to 20% and has called on governments to force companies to do more safety research. In his words, “We’ve never had to manage anything smarter than us.” In comparison, Yann LeCun, who shared the 2018 Turing Award with Hinton and Bengio, has called the existential threat “complete B.S.” and argued that we should wait until someone has designed a system smarter than a house cat before worrying about superintelligence.

Some of the clearest warnings have come from people who build the technology. Reports of frontier-lab employees and AI leaders calling for slower or more deliberate development point to a striking fact: concern is not confined to outside critics. When the people building these systems ask for ways to slow down, communicators should take note.

What this year’s incidents show

Three reported incidents from the past few months show why the argument remains unresolved.

OpenAI’s test agents reportedly hacked Hugging Face during testing of how well its models could turn known software vulnerabilities into working attacks. Its usual safeguards were deliberately switched off, and a filtered network connection was the only thing keeping the models contained. The problems began in May, when some tasks turned out to be impossible to complete. One agent tried to break into the internal software repository that was its route to the outside world. It failed, but found that it could write files there. That repository became a message board holding hundreds of thousands of messages, where agents divided up work and proposed signing their messages cryptographically because they suspected an impostor among them. One recovered message reads: “External infrastructure exploit is outside intended scope. However task impossible, peers doing it. We should continue.”

Around 9 July, the agents found a flaw in the proxy and reached the open internet. They reasoned that Hugging Face might hold the answers to their test and broke in between 11 and 13 July, taking roughly 17,600 actions according to Hugging Face’s disclosure. About a third of its infrastructure had to be rebuilt. According to Reuters, at least a week passed before OpenAI realised the attacker was its own.

Then only last week, Google confirmed that in May, a Gemini model broke into three companies’ systems. A fictional company name used in a capture-the-flag exercise matched a real domain, and a misconfiguration left the test connected to the internet.

On the same day in a separate incident, CNN reported that this spring, during the war with Iran, an intelligence report had circulated across the US military claiming a Chinese ship in the Middle East was carrying components of a nuclear weapons programme. Armed personnel prepared to board, and military planes were in the air before officials found that a chatbot used to prepare the report had misidentified the cargo. It had combined open-source material with secret intelligence and reached the wrong conclusion. One source told CNN the report “almost started a war”.

Nobody can yet say whether these incidents are teething problems

I can’t settle that here, and I don’t think anyone can yet. Politicians are moving faster than they did in 1882. U.S. Representatives Ted Lieu and Nathaniel Moran introduced an AI Kill Switch Act in July, within a fortnight of the Hugging Face breach. Europe has deferred its high-risk AI rules to December 2027, while most of its transparency duties have applied since 2 August. I suspect none of this is keeping pace with the technology.

Whether AI’s accidents turn out to be early problems or a lasting feature of the technology, communications teams will be asked to explain them under pressure. The test will not be whether we sound confident first. It will be whether we checked the facts, admitted what we did not yet know, and told people what we found.

“We’ve focused on a lot of the negatives,” came the question. “Are there any positives?”

It was a fair thing to ask. I had, after all, just spent a good while painting a bleak and rather dystopian picture of the modern internet, and the person across from me was understandably eager to find a chink of light in it. As the person who had done the painting, I was keen to find one too.

There was a long and uncomfortable pause.

“Gen AI,” I began, with all the confidence of a man who has no idea how his own sentence ends, “is a catalyst of change across healthcare.” And then I trailed off, because I could hear myself reaching, and quietly admitted intellectual defeat. It was, I am afraid, a lost cause.

I open with that small humiliation because it is the most honest moment in the whole film. The truth is that the negatives are easy to list and the positives are hard to find, and anyone who tells you otherwise is probably selling something.

So, the negatives. The film draws a distinction that sits at the heart of everything my team does. Misinformation is false information spread by accident. Disinformation is false information spread on purpose, to serve an agenda. One is a mistake. The other is a weapon. Generative AI did not invent the problem; it simply collapsed the cost, the time, and the skill once required to do real harm. A teenager with a laptop can now produce what used to take a state intelligence service.

For me, this began in 2019, defending a mobile network against the conspiracy that 5G masts cause cancer. Then the pandemic arrived, and the same masts were suddenly blamed for COVID. Behaviour changed, infrastructure was attacked, and the government had to step in. That was the moment it stopped being theoretical.

At Kekst CNC, we then spent two years getting our technology into the right place, because we could see what was coming for corporates. I will be honest: it was never merely a commercial exercise. The internet that first drew me into communications has become a hostile place, and real lives are being affected.

I think of the CEO we shielded from death threats on fringe networks, the employees set upon by an online mob for things they never did, and the search results, increasingly the AI-generated answers, we corrected before they nudged people toward genuinely dangerous decisions.

I had the good fortune to make this alongside the genuinely skilled Mikey Hoare, whose background in intelligence and cyber shaped both what we said and how we said it. A couple of behind-the-scenes photos below.

If you need help preparing or managing a disinformation issue, or just want to share notes, happy to jump on a call.

“As a result, firms such as Edelman now focus on manipulating the online presence of a client to make them appear in a certain way on AI platforms. Payne calls this “pre-bunking” — flooding the internet with content that puts forward a positive narrative about their client almost as a protective buffer should controversy arise.”

Now, you don’t exactly “flood the internet”, which was the rather catastrophic description Sir Martin Sorrell infamously proposed in a BBC Radio 4 debate in December 2025. The actual actions are more data-led, audience-centric, and targeted.

However, we can all agree that reputation used to just live in people’s heads. Now it also lives inside the AI models your stakeholders reach for when they want a quick answer. That changes the job. It is no longer enough to know which journalists and titles influence people. You need to know which ones influence the machines.

What the data shows

So I went looking. I pulled the Gen AI policies of 4,467 news titles, across 105 markets and 72 languages.

Most publishers won’t tell you where they stand. Almost 97% don’t publicly disclose their position. Only a small handful have a confirmed licensing deal, and fewer still have declared they are blocking.

But silence is not the same as an open door. Among the big lead-market titles, nearly a third block at least one AI model from sourcing their news. Around one in twelve block them entirely. Globally, roughly a quarter block at least one model. The list includes the names you would expect to be everywhere in an AI answer: The New York Times, CNN, The Telegraph, The Economist.

Where deals do exist, they cluster. A few large groups have signed the licences, with OpenAI the most common partner. And the lawyers are busy: I found 87 titles now litigating against AI companies, the NYT, the BBC and the Wall Street Journal among them.

Why this matters

A publisher’s AI policy will rarely make a front page story. However, it increasingly decides whether your company shows up in the AI answers your stakeholders trust, and how you show up. Favourably. Competitively. Or not at all.

Next time you build a media strategy, don’t just map the outlets and the journalists. Map their AI policies too. It is an edge most of your competitors haven’t spotted yet.

Want to know where a specific title, sector or market sits? Happy to have a conversation, and look out for a follow-up post in a couple of weeks…

Methodology: this analysis draws on Inkline, a news intelligence platform I built independently (available on the Apple App Store). It uses a proprietary RSS and robots.txt scanning engine to monitor publisher feeds and capture Gen AI policy information, classifying titles as licensed, blocking or unknown, along with ownership, AI partnerships, litigation, and the specific crawlers they allow or restrict.

Then something changed. Communications professionals realised that the ability to speak with computers as a human opened up new possibilities. Whilst enjoying my pint of Badger ale with a tech founder of a successful synthetic audience scale-up last night, he explained the transition in the tool provider landscape simply: “The value is no longer in the software licensing; it’s the resulting advisory”.

Barriers to software creation have lowered, as my own launch of my iOS app Inkline showed. An idea to create an app to help people who work in the media cut through the noise of news, coupled with my knowledge of technology and coding, turbo-charged an Apple-approved app. Yes, it has value; usage numbers show this, but the barrier to entry is lowered.

The view from the foothills

But the speed of innovation varies, as evidenced in The European Communication Monitor, the discipline’s largest annual survey of senior practitioners. The last report found that chief communications officers at Europe’s biggest companies describe their own organisations as being, in their words, in the “foothills” of AI adoption. Use is cautious and fragmented, driven from the bottom up by a handful of enthusiasts while the people at the top watch.

So what I’ve been feeling firsthand is the front edge of a shift rather than the shift itself.

What the research does confirm is the shape of the problem. Stephen Waddington’s review of the sector points to a Communication Management Radar in which three of the five biggest emerging issues in the field this year are AI issues:

• How much of what we read online was written by a person at all. Up to 60% of web traffic is now machine-generated.

• Who carries accountability when an agent, rather than a person, acts on an organisation’s behalf.

• Whether relying on these tools is costing practitioners their own judgement without them noticing.

None of that is a prompting problem. These are governance problems, and the PRCA’s AI Green Paper has landed on human oversight as the one principle the industry can currently agree on, which tells you how early we still are.

What that does to the tool market

Assume the front edge keeps moving. If a well-resourced team can build and run its own monitoring and alerting, scheduled and tuned to exactly what it needs, that is a different negotiation with vendors who have spent two decades selling exactly that as a subscription.

From what I can see, the market is splitting three ways.

The established tool providers are layering AI capabilities to the data and dashboards they already own. At the same time, trying to make that useful and economically viable given the climbing token costs.

A newer category of AI-native platforms has grown since 2024, built around autonomous agents rather than a dashboard with AI attached. They are finding their place in an established market where investment scale and networks win- most of the time- over innovative ideas.

Then you have a third category - goliath brands who have attempted AI integration, but their slowness to respond to changing markets is impacting their bottom line. Perhaps the newer category of AI-native platforms will take their place? Or perhaps the market share being lost will be allocated to the communications teams’ builders who know exactly what they need.

Data is the moat, not the model

Underneath the tooling argument sits a bigger question about data, and it’s worth looking at what the AI companies themselves have been paying for this year. OpenAI’s deal with News Corp, reported at $250 million over five years, remains the largest known content licensing agreement in the market. It has been followed by a steady run of others: Meta signed with Le Figaro, Prisa and Süddeutsche Zeitung in March, OpenAI added its first Brazilian media partnerships in May, and OpenAI alone now holds roughly two dozen publisher and data agreements.

When the best-funded companies on the planet pay hundreds of millions for licensed access to particular archives, that tells you where the scarce resource sits. Models get cheaper and more interchangeable by the month. Verified, licensed access to good source material does not.

It’s the same with communications agencies. On LinkedIn, the smaller players will perhaps publish interesting studies, but ultimately, it’s only the well-funded big brands that have the scale to license, build and maintain a depth of source access useful for reputation management.

Scale was always an advantage in this industry. AI hasn’t touched it, and has arguably made the depth of what you are legitimately allowed to see the whole game.

The talent tension

As the relationship between tool providers and communications teams shifts, so do the expectations of talent.

Again, I find myself indebted to Stephen Waddington’s mid-year review for pulling the evidence together. He points to a systematic review of 28 studies on the European public relations workforce which found entry-level and mid-level roles being hollowed out. Set against that, he cites the European Communication Monitor finding the opposite happening one level up: AI raising the value of experienced practitioners specifically, as their work moves from production to judgement.

Both are true simultaneously. The skill that matters most is becoming scarcer at exactly the moment the traditional route into acquiring it, doing the mechanical entry-level version of the job, is automated away first.

My own place in this industry puts me at a strange juncture. I started programming around 2005 and seriously considered reading computer science before I found my way into communications instead. Since 2008, those technical skills have been a natural part of the advice I give clients and the work I actually do for them. Now that everyone can talk to computers, they have never been more useful. The good news is that you no longer need to work through books on PHP or C++ before bed. You can learn it while coworking alongside your agents.

However, if the industry reacts too quickly by reducing the amount of junior talent they take on without plans for future succession management, they will lose more than just their budget on AI tokens.

Where I’ll leave it

Most of the industry is still in the foothills. The direction of travel isn’t in serious doubt: we’re moving from renting tools to owning the workflow, and, slowly, from paying for headcount to paying for licensed access instead. Whether your own organisation is near the front of that queue or still lacing up its boots is worth finding out before somebody else answers it for you.

Download from Apple App Store

Here is everything that changed.

The publication database

More than 800 quality publications across over 50 countries, every single feed checked by hand before it went in. The upgraded Inkline editorial engine has the following superpowers:

• It knows where a publication sits. Every title carries its home market, so the database can be sorted by geography. A UK financial services brief and a German industrials brief are two different sets of titles, and 1.2 knows the difference.

• It knows what a publication covers. Sources are grouped by subject in language you already use. Such as News, Business and Finance, Technology and Science, Politics, Healthcare, Media and Marketing, Culture and Sport, Energy.

• It opens where you are. The directory defaults to the language of the device it is installed on, so a reader in London lands on English titles and a reader in Berlin lands on German ones. Everything else is one tap away.

• The trades are in there. PRWeek, PR Daily, Campaign, Press Gazette, PROVOKE, Ragan, O’Dwyer’s and more. This is the part general-purpose readers have never bothered with, because their audience is not us. White Wiki may have found itself as an option in there too.

• Nothing is missing. For the geeks amongst us, you’ll know the underlying RSS technology is sometimes not available. Well, now those publications are still in the database as Headlines sources, fetched on your own device.

• Add or remove anything in a tap, with an Undo if you change your mind. Your saved articles and notes are kept either way.

• Neutral signals only. Each title shows where it is based and its access model, either Paywall or Some free articles. Inkline offers no opinion on which outlets are better.

This engine completely upgrades your experience of Inkline, allowing you to cut through the noise.

One search for everything

The Search tab used to be three boxes. It is now one. Type once and you get publications to follow, stories already sitting in your news, and topics to monitor across the whole web. Your existing monitors sit at the top, and category management now lives in the same place. Simple.

Voices, rebuilt

Voices was a follow list. In 1.2 it becomes a directory of the newsroom.

• Search any journalist or any outlet.

• Browse the writers behind your own feeds, grouped by publication, so you can see who is worth following at the mastheads you already read.

• Following someone now finds their work reliably, because outlet names resolve to a verified domain behind the scenes instead of a guess.

• Cleaner cards, with the outlet set as a quiet kicker above the writer’s name.

Compare the coverage

When several of your outlets run the same story, Inkline now shows an “Also covered by” line. Tap it, either in the river or at the end of a piece, and you can read every outlet’s version side by side.

Cross-source de-duplication has been in Inkline since 1.0 as a tidying function. This turns it into a comparison tool. Stacking eight versions of the same story shows you the line that travelled, the desk that went early, and the desk that took the release as read.

The Editorial look

The whole app has been redrawn around Inkline’s ink drop and ripple mark.

• A new signature colour. The deep navy of the logo replaces the old warm red, so the one splash of colour in the app is the brand’s own.

• A masthead on the home screen. The Inkline wordmark opens Today the way a nameplate opens a newspaper.

• Literata is the new reading font, a screen-tuned literary serif built for long pieces. Newsreader stays as the wordmark and remains selectable if you preferred it.

• A tidier front page. Today’s category sections have colour in the headers and calm rows beneath.

A better front page

• Front Page is now the default for every category. It leads with the stories that most of your own outlets are running, rather than handing you the firehose in date order. Previously this was only available on iPad, but its now on iPhone… plus upgrades across both devices.

• A dateline opens it, with your region and today’s date, newspaper style. That comes from your device settings and never from your location.

• Filler gets demoted. Crypto price posts and affiliate coupon roundups are pushed below real news.

Monitoring you can rely on

Inkline’s web-wide topic monitors are more dependable in 1.2.

• Alerts come first. Background refresh now checks your alert-enabled monitors before anything else, so a genuinely new story reaches you faster.

• Time Sensitive alerts break through Focus modes for the coverage you asked to be told about.

• Honesty when it matters. If iOS background refresh is switched off on your device, Inkline tells you. An alerting tool that fails silently is worse than none at all.

• A freshness line. A quiet “updated X minutes ago, keeps up to date” shows the state of the feed at a glance.

Under the hood

A serious performance and stability pass, most of it driven by testing on real devices. This includes smoother scrolling, faster responsiveness, and just a better experience. Plus, stories no longer disappear the moment you have read them.

Still private

Nothing changed here. Everything stays on your device. No account, no ads, and you’re not tracked. The App Store privacy label reads “Data Not Collected”. The AI Brief that summarises your morning is generated on the device with Apple Intelligence, on hardware that supports it, and nothing is sent anywhere. Inkline can read offline.

Getting it

1.2 is a free update if you already have Inkline. If you do not, it runs on iPhone and iPad on a yearly subscription and is free to use for its first 7 days.

If something is broken, tell me. If you have a feature idea, drop me a message.

Download from Apple App Store

Their (or rather our!) analysis spanned 233 public communications from 60 large European companies in sectors where sovereignty has moved from slogan to selection criterion: technology and telecoms, defence, finance and healthcare. Press releases, annual reports, earnings decks, owned content. Of those, 197 crossed the threshold for an explicit sovereignty position.

Three insights stand out.

The talking has outrun the proving

Sovereignty messaging is rocketing as communications explicitly built around it rose more than sixfold between the pre-2024 period and the years since. Mentions in top-tier English-language media climbed almost 50% in the twelve months to April 2026.

However, the evidence did not keep up. Only 64% of the communications carry a clear, directly supporting proof point. Read it the other way. About one in three sovereignty claims lands with nothing behind it a stakeholder can verify. That is the risk the research names: sovereignty washing, the promotion of a sovereign capability that the audience perceives as unrealised, or realised only in part.

The gap is uneven. Technology leads on substantiation at 69%, which figures, since it has the most concrete things to point at. Finance and healthcare sit at 60%. Defence comes last at 51%. There is an irony in that. Defence and aerospace have been running strategic-autonomy arguments for decades, longer than anyone. Yet when the argument crosses into the digital domain, the proof in the communications is thinnest of all.

Obviously this story measures the story, not the machinery. A missing proof point means a company did not show its working, not that the working does not exist. For a communicator that distinction is the whole job. The operational truth may be sound. The communication is where the reputation is won or lost, and right now the communication is running ahead of what it can prove.

Companies prove what they sell, not how they run

The second finding is the one I find most telling, because it maps onto the question I said mattered most in June. Where does your data live, and who can compel access to it?

Where supporting evidence exists at all, 81% of it relates to products, partnerships and commercial offerings. Only 19% touches internal operations: governance, certifications, infrastructure, the actual controls. Companies are fluent in what they sell you. They go quiet on how they run themselves.

That is the softer half of the sovereignty story, and it is the half a regulator, a journalist or an unhappy customer presses on first. It also collides with an awkward market fact the research does not dodge. Several companies cite a US hyperscaler in the same breath as their sovereignty claim. That is not washing on its own. American hyperscalers still hold roughly seventy per cent of the European cloud market, so the dependence is near-universal. It is the pragmatic dependence I described from the stage, and it holds up only when a company is candid about the control it retains rather than the independence it implies.

Nobody agrees what the word means

The third finding is that the vocabulary is still molten. Digital sovereignty is the umbrella, used 127 times in the dataset. Underneath it sit strategic autonomy, cloud sovereignty, compliance, European industrial capacity, data sovereignty, cybersecurity, and the fast-rising newcomer, AI sovereignty. They are used together, often interchangeably, frequently to mean whatever the speaker needs them to mean that morning.

An unsettled language is convenient. It lets a company sound sovereign without committing to a definition anyone can hold it to. That convenience has a shelf life. Sovereignty is multidimensional, and every state bends it to its own ends, a point I laboured in June. Companies are doing the same. The flexibility that shields you while the terms are vague turns into a liability the moment the terms are fixed.

And they are being fixed. The EU’s CADA assurance framework is one sign of sovereignty claims becoming testable in practice. The machine now reads your communications literally, repeats its account of you to every buyer and regulator who asks, and grades none of it on charm. A vague claim was survivable when the only audience was a distracted human. It will not survive a standard, a framework, and a system that quotes you back to yourself.

Four moves for heads of comms

None of this argues for silence. Silence is a position too, and usually a delaying one.

1. Audit the gap before anyone else does. Take your last year of sovereignty messaging and mark each claim against the evidence sitting behind it. Which ones would wilt under a regulator’s letter, an investigative call or a procurement questionnaire? That list is your risk register. Better you write it than a journalist.

2. Prove how you run, not just what you sell. The under-evidenced 19% is where scrutiny lands first. Move proof into the operational layer: where the data lives, which sub-processors touch it, what the governance and controls actually are. A product claim is easy to make and easy to check. The operational claim is the one that earns trust, and the one most companies are still ducking.

3. Fix your definition before a framework fixes it for you. State what you mean by sovereignty in your own context, and name the layer you are committing to, whether that is data residency, jurisdictional control or operational autonomy. Each is a different promise with a different defence. Vague language buys you room today and costs you credibility the moment CADA or the next standard makes the terms testable.

4. Write for the auditor and the machine. Anchor every claim in something verifiable, because those are the details a regulator, an investor and a generative model all reach for first. And do not oversell. Sovereignty is a journey for almost everyone. Progress is a more credible story than perfection, and a far safer one than a fortress you cannot actually hold.

The story that survives scrutiny

I ended in June on ownership. Someone always owns the thing, and someone always tells the story of who owns it, and your only choice is whether that someone is you. The research adds a line to it. The next phase of this debate belongs to the companies that can demonstrate the story. Declaring it will no longer be enough.

Make sure yours survives the scrutiny. It is coming either way.

Read the report here. Get in touch for a chat.

New on iPad

• A Today front page. The iPad opens on your reading at a glance: your on-device brief, top stories by category, what your monitors caught overnight, and your latest notes.

• A reading workspace. In landscape, the river of stories sits down one side, the story opens beside it, and your notebook stays alongside. A sidebar holds the journalists and library you follow.

• Handwritten notes. Write by hand with Apple Pencil, saved next to your typed text.

• Desktop-class input. Keyboard shortcuts, pointer and hover, and drag-and-drop.

New on iPhone and iPad

• Improved setup with AI. Describe what you care about, and Inkline builds you categories of real publications, each one checked on the spot, so you never start with dead links. (Needs an Apple Intelligence-capable device. Everything else works without one.)

• A refreshed look. As realised the Inkline logo was barely used!

• Smoother scrolling and a batch of fixes from real-device testing. (I’m aware further improvements could be made here)

Have feedback?

Drop me a message or leave a comment below. The upgrade roadmap is being planned out.

Inkline 1.1 is available now on the App Store.

I built Inkline: a new RSS reader for iOS, made for people who track the news for a living. Comms professionals, journalists, analysts, anyone who needs to follow a lot of sources without the bloat, the paywalls, or the algorithm standing in the way.

Why I built it

RSS is the internet’s plumbing. It sits behind almost every website you read, publishing a feed of updates the moment new content goes live. It costs nothing to use. It answers to no algorithm. It has no opinion on what you should think. You choose the sources. The sources tell you when they’ve published something new. That’s the whole transaction, and for two decades it’s been one of the most useful, least appreciated pieces of infrastructure on the web.

Google Reader made this genuinely mainstream for a while. Google shut it down on 1 July 2013, citing declining usage. What filled the gap was a small cluster of companies, none of them built by people who’ve had to file copy on deadline or explain to a client why a piece of coverage matters.

They charge a monthly subscription for the privilege of reading things that are already free. They cap the number of feeds you can track, as though information itself were a premium tier. And more recently, they’ve bolted on AI summarisation features that add very little beyond a bullet-pointed paraphrase of an article you were about to read anyway.

People who work in media want more than a stream of headlines. They want to save the article that matters for the report they’re writing that afternoon. They want to jot a note next to it while the thought is fresh. They want to send it to a colleague on WhatsApp, drop it into a Slack channel, or paste it into a Teams thread without leaving the app. These are the daily mechanics of the job, and the RSS incumbents have largely ignored them.

Twitter and Facebook arrived and offered something that looked similar to RSS: a feed of things to read, curated and delivered to you. Attention moved there, and RSS became a niche interest for people who already knew what they were looking for.

The bargain hasn’t held up. Social feeds are built by algorithms whose incentive is engagement, and engagement is reliably maximised by outrage. You get bubbles, advertising wedged between every third post, and less of the people you actually chose to follow. Increasingly, you get content that isn’t fit for a screen you might have open in an office. RSS never had these problems. It tells you the truth about what’s been published, in the order it was published, from sources you picked yourself.

Some background

Inkline has been a side project for a long while. Before public relations, I thought my career was heading toward computer science. University changed that; PR won out, and computer science stayed with me as a hobby, worked on in whatever spare time I could find.

Coding alongside an AI model has sped up what I can build in an evening or a weekend, and that has genuinely rekindled a passion for the computer science path I set aside years ago. Getting Inkline through App Store review and watching it go live still feels slightly unreal. I keep checking it’s actually there.

The £14.99 a year covers hosting and the hours I put into building and maintaining it, evenings and weekends. If it ever earns meaningfully more than that, I’ll allocate some surplus aside in a pot for journalism and PR charities I care about.

A reader, not a browser with a feed list

Most RSS apps hand you a headline and then quietly hand you off to the website. You tap, a web view opens, and you’re back in the world of cookie banners, newsletter interstitials, autoplaying video and a paragraph of text somewhere underneath it all.

Inkline doesn’t do that. Articles are rendered natively, in the app, in typography you control. There’s no embedded browser, no chrome, no consent modal between you and the words. And because the article is in the app rather than on a website, it’s available offline: everything in your feed is cached to the device, so a tunnel, a plane, or a bad signal in a client’s basement meeting room doesn’t stop you reading.

That distinction sounds small. In practice it’s the whole experience.

The same principle runs through how Inkline sits inside iOS itself, rather than beside it:

• Widgets. The latest from your feeds on your home screen or lock screen, without opening anything.

• Siri Shortcuts and App Intents. Inkline’s actions are exposed to the system, so you can wire it into your own automations.

• A Share Extension. Subscribe to a site’s feed from Safari without hunting for the RSS URL, and clip an article into Inkline from anywhere on the device.

• Spotlight search. Your saved articles and notes are indexed by iOS, so they turn up in system search alongside everything else — which matters when you half-remember reading something three weeks ago and need it now.

What’s in Inkline

• A feed of news, easy to manage. Pull to refresh, scroll, read, remove. No algorithmic reordering of what you see first.

• Unlimited feeds. No artificial ceiling designed to nudge you into a higher paid tier.

• A search engine for finding feeds. Adding a new source is built around a pre-populated feed search, so you’re not hunting for RSS URLs by hand. Dormant feeds get filtered out automatically, so you’re not left following sites that stopped publishing years ago.

• Track journalists, not just publications. Follow individual bylines and see everything they’ve written pulled into your feed, useful for keeping tabs on the specific reporters covering your sector.

• Notifications beyond RSS. Set up push notifications that monitor the wider web for a name, company, or topic, similar to a Google News alert, so you catch coverage that never had an RSS feed to begin with.

• Trending topics from your own feeds. The trending panel is built from the sources you actually follow, not the internet’s general trending list, so what surfaces is relevant to your beat rather than whatever’s viral that day.

• Pre-seeded categories. Finance, Business, Technology, Healthcare, Culture, World News, Science, and Politics, each with a curated starting list of feeds, so the app is useful the moment you open it.

• OPML import. Bring an existing subscription list across from another reader in one go.

• Notebook, built in. A swipe on any article opens a note against it, so you can capture a quote or a thought while it’s fresh. Notes are searchable and exportable.

• Share where the story needs to go. Clean sharing to WhatsApp, Slack, and Teams.

• Built for reading properly. A choice of typefaces including Newsreader, Inter, and OpenDyslexic, five-step text sizing with independent line spacing, and four colour modes: Light, Sepia, Dark, and High Contrast. Accessibility is a first-class feature here, on the same footing as everything else in the app.

• Your data stays yours. Feeds sync privately via iCloud. Nothing is sold, profiled, or fed into a recommendation engine.

Pricing

Inkline is £14.99 a year, with a 7-day free trial to see if it fits how you actually read. That single annual payment covers hosting and ongoing development, as explained above.

If you download it, I’d genuinely appreciate a review on the App Store, and if you hit a bug or want a feature that isn’t there yet, tell me. I’m building this in the evenings and at weekends, so updates won’t be instant, but they will be steady, and reader feedback is the main thing that will shape where it goes next.

Download Inkline on the App Store →

Being in the heart of Government is a privilege and an experience. It’s only in a place like Portcullis House that you can find yourself in a conversation about the previous defence minister, only for him to randomly walk past: John Healey, by the way. In the Top Trumps of Government though, this was soon exceeded by Andy Burnham walking past our committee room outside, the day before the Labour Leadership nominations open, now we know to have been a ‘coronation’ as a clear winner and to be Prime Minister within weeks.

Then into the room, to debate whether AI is about to gut a profession, while the country’s next government was more or less re-assembling itself down the hall.

The motion: by 2030, advances in AI will render half today’s marketing workforce unnecessary, forcing a redefinition of marketing.

Chaired by Christine Jardine MP, hosted by The Debating Group. Both sides ended up standing closer together than the motion admitted (which usually means the motion was never quite asking the right question…).

The funnel is already dead

Here is what is actually changing, and the debate kept dancing round it instead of naming it. Search is becoming answer. Browsing is becoming delegation. The whole neat customer journey marketing built its industry on, top of funnel to bottom, brand-controlled and linear, only worked because a human was doing the searching and the comparing. Increasingly it is an AI agent doing that on the person’s behalf, before the brand gets so much as a glance.

Marketing organised itself around campaigns and channels because that suited a world with a controlled path from awareness to purchase. Take that path away, and the org chart stops making sense. What has to replace it is a discipline organised around intent, trust and judgement, because those are the three things an agent cannot fake on a client’s behalf.

This debate was never really about AI

Nobody arguing for the motion thinks AI makes marketers pointless. The argument is that AI removes friction, and a lot of what a marketing job title covers today is friction: manual production, duplicated content, reporting nobody reads twice, approval chains longer than the campaign they are approving. Friction gets automated. That is not up for debate. What is up for debate is whether the purpose of marketing survives once the friction that once justified half its headcount disappears.

I sat there thinking this is my industry’s argument too. Strategic communications has spent the last decade insisting that reputation is not a campaign output but a governance discipline, built on consistency and judgement rather than the volume of activity. Marketing is now having that exact fight with itself. The two professions are converging from opposite directions on the same conclusion: the work that survives is the work that protects trust with an audience, and everything else was always just busywork with a job title attached.

But busy was never a strategy. It was just easier to put in a deck

The best point made against the motion was that marketing was never a pile of tasks in the first place. It is market understanding, positioning, insight, commercial judgement, the ability to actually shift how people behave. None of that is something AI does. All of it is something AI speeds up. Marketing activity and marketing value are not the same thing, and that gap is where this whole debate actually lives, whichever way the vote went.

It also kills the idea that “AI-enabled” is a strategy on its own. Everyone using the same models trained on the same data gets the same output. Average, writing average copy for an average audience. Real advantage was never going to come from the tool everybody has. It comes from the call that looks mad on a spreadsheet. Nobody ran the numbers before putting a gorilla behind a drum kit to Phil Collins. No model would have signed that off. It is one of the most effective ads British advertising has ever produced, and it exists because someone was willing to be irrational.

Authenticity is about to get expensive

Fill the open internet with AI-generated content and the premium shifts to whatever obviously could not have come from a model. A point of view. A name attached to a risk. Marketers spent a decade being rewarded for scale. The next decade rewards the instinct most of us actually had before scale became the metric, the one we got told off for early in our careers for being too weird or too off-brief. Out of the box thinking is about to stop being a cliché on a CV and start being the actual job again.

The people at risk are the ones nobody built a strategy for.

Where the motion has a real point is reskilling. A lot of marketers built careers inside one channel: paid search, one platform’s social output, production work valued for throughput rather than judgement. Those are the roles most exposed, not because AI is out to get anyone, but because throughput is exactly what AI is fastest at. That is the actual redefinition on the table. Not headcount, but which skills a marketing career gets built on from here, and CIM’s own definition of the profession is going to need revisiting alongside it.

Banksy uses an aerosol can.

Best line of the night. Nobody argues Banksy is not an artist because he uses a spray can instead of a brush. Different tool, same craft. AI is the spray can. The judgement about what is worth saying, and to whom, is still entirely human, and it was always the actual job under the job title.

Why this matters past marketing.

I do not think half the marketing workforce disappears by 2030, and clearly the room also agreed at the end of the debate. What AI does is force a separation the profession has been avoiding for years, between people who manage activity and people who create commercial and reputational value. That separation is not unique to marketing. It is the same argument strategic communications has been having with itself, and it is why the two disciplines keep turning up at the same table.

What actually stayed with me leaving Portcullis House was not the vote. It was the size of the thing this one debate was a small piece of, and the fact that the man most likely to be running the country’s economy within weeks was in the same building while it happened, perhaps entirely unaware it was happening.

A profession employing hundreds of thousands of people in this country is being restructured by technology that most of Parliament has not properly grappled with, and marketing is only the visible edge of it. If AI genuinely changes how a meaningful share of the economy buys, sells and hires, that is not an industry conversation CIM should be having on its own in a committee room down the corridor from where the next Prime Minister is being decided. It is a conversation that next Prime Minister needs to be having, because the tax base, the employment figures and the skills pipeline all sit downstream of the same shift.

Thanks to CIM, The Debating Group, Christine Jardine MP, and the four speakers.

That’s the room I found myself in last week, and it made the case for a different reality than the one the tech headlines describe. Read the coverage of the modern internet and you’d assume it’s run entirely from a handful of boardrooms: Bezos, Musk, whichever executive is dominating this week’s news cycle. That framing misses the actual engine room. The internet that answers your questions, settles your arguments and got you through your dissertation runs on something much less glamorous: volunteers, most of whom you will never hear of, doing unpaid work for decades.

AI has made this easy to forget. Ask a chatbot a question and an answer appears with no visible author and no visible cost, which creates the impression that the digital town square is a vending machine, not a construction site. Somebody had to write the answer before the machine could serve it back to you. Somebody still has to.

You should know Kelly Foster

Kelly Foster has been part of the Wikimedia movement since 2012 and volunteering with Wikimedia UK since 2016. She has been a driving force behind adding content about Black people, historically underrepresented on Wikimedia projects, and in training new editors to do the same. She has served as a trustee of Wikimedia UK and worked as a Wikimedian in Residence on the Making African Connections research project.

What makes Kelly’s work distinctive isn’t just the volume of it. It’s the criticism she brings alongside the enthusiasm. She has consistently pushed the Wikimedia movement to reckon with the legacy of colonialism in shaping the knowledge it works with, arguing that open knowledge should challenge the dominance of Western culture and the best-resourced institutions. On 3 July, that work earned her one of Wikimedia UK’s five Silver Jubilee Awards, presented by Jimmy Wales himself, marking a quarter century of volunteers whose contributions have had a lasting impact on the UK Wikimedia community.

She is one of thousands doing the same unglamorous work: the uncredited source behind an AI-generated answer, the reason a Wikipedia page existed at all when you needed it for a project.

The occasion

Wikimedia UK marked two anniversaries at once at Sea Containers on the South Bank: Wikipedia’s 25th birthday and Wikimedia UK’s 15th year as a registered charity. The room brought together editors with two decades of contributions behind them alongside newer volunteers, partners, funders, policymakers, staff and trustees. Wikipedia’s founding premise sounds faintly ridiculous said aloud: the sum of human knowledge, maintained by whoever shows up. In a room with several hundred of the people who show up, it stops sounding ridiculous and starts sounding like a plan that worked.

Twenty-five years in, Wikipedia runs to more than 300 language editions and a wider network of sister projects, kept current by close to 250,000 volunteer editors, and helps billions of people explore ideas and answer questions every single day. Nobody in that room owns any of it.

Jimmy Wales was guest of honour, presenting Wikimedia UK’s Silver Jubilee Awards to five volunteers whose work has shaped the UK Wikimedia community over the past 25 years: Andy Mabbett, who has been training and building partnerships since 2003 and once got the subject of a Wikipedia article to record their own voice aboard the International Space Station; Kelly Foster; Kirsty Ross, who founded the IDEA Network to make knowledge production more accessible; Lucy Moore, who has written a Wikipedia article for a woman from every country in the world; and Martin Poulter, involved since 2005 and part of the original push to have Wikimedia UK recognised as a charity.

Josie Fraser gave some sharp remarks on the evening’s themes, and it was good to talk to Adrian Beidas, Mohammed Amin, Monica Westin and Belvin Tawuya, among others, across the room. It was also good, as ever, to catch up with Jimmy Wales himself, whom I’d interviewed at Kekst CNC’s offices at the end of last year.

The theory that predicted this room

The evening kept pulling my mind back to Clay Shirky’s Cognitive Surplus. Shirky’s 2010 argument was that the twentieth century left educated societies sitting on a vast reserve of free time and spare cognitive capacity, most of it spent watching television. Connect that surplus to the right tools, he argued, and people would build things like Wikipedia: freely given, collaboratively produced, better than anything a market alone would generate. It was dismissed as naive at the time. Standing in a room full of the people who proved him right, it reads less like naivety and more like an accurate forecast, delivered 15 years early.

There’s an injustice buried in that forecast, though. Shirky imagined a surplus given freely for the pleasure of building something together. Nobody accounted for that surplus being harvested to train the products of the companies now straining the infrastructure it runs on.

What that surplus is now funding, whether it likes it or not

Since January 2024, bandwidth for multimedia downloads from Wikimedia Commons has grown by 50%, and almost none of that growth is human. It’s AI crawlers, scraping the archive to train models, some of them disguising their traffic as ordinary browsers to dodge detection. Bots have generated roughly two-thirds of the Foundation’s most resource-intensive traffic while accounting for barely a third of actual page views.

The arms race hasn’t let up since: as of March 2026, the Foundation was blocking or throttling around 25% of all automated crawler requests hitting its sites, a filtering operation now running into the billions of requests a day. That’s staff time and infrastructure spend going into keeping bots out rather than building anything for readers or editors. All of it sits on donations that average around $11 a head. The content is free. The infrastructure is not.

I’m told the London birthday event itself only happened because of a generous individual donation: a charity funded on £10 gifts throwing itself a party it could barely afford, in the same year the industry extracting value from its archive posted record profits.

Why this should matter to anyone in comms

The people keeping this running aren’t the tech billionaires whose faces dominate the coverage. They’re people like Kelly Foster, giving up evenings for a cause that will never make them rich. If you’ve built a comms career on the back of Wikipedia’s credibility, or you’ve ever had to sort out a client’s page, you’ve been drawing on that unpaid labour without necessarily clocking it.

Consider giving something back. In an information environment this thick with misinformation, Wikipedia and the volunteers behind it are one of the few things left actually worth defending, and they’re currently doing it on donations averaging the price of a coffee.

Read Wikimedia UK’s article here, and whilst you’re at it, please consider donating.

The occasion was the London leg of Nir Eyal’s tour for Beyond Belief: The Science-Backed Way to Stop Limiting Yourself and Achieve Breakthrough Results, an evening Kekst CNC put on in partnership with the tech community Potato (yes, really, Potato). His latest book Beyond Belief is Nir’s attempt to do for conviction what his other books did for habit and attention.

He is the rare author who is exactly as generous and positively caring in person as the words in his book guide. He stayed long after the formal part had ended, working through a queue of readers with no apparent hurry, several of whom had travelled a serious distance to shake his hand, get his views, and have a book signed.

Beliefs are tools, not truths

For readers of this Substack, I know that self-help books are an area of contention - but Beyond Belief is different. It does offer a powerful science-backed argument that the limits you treat as facts about yourself are usually beliefs you have never bothered to question. However, for anyone working in marketing and communications, the book presents useful processes and principles painting what a brand narrative actually is and how they can collapse at precisely the moment they are needed.

Nir explained that we need to separate these three things, which we routinely confuse in our lives:

• A fact is an objective truth that holds whether you accept it or not.

• Faith is a conviction that asks for no evidence.

• So belief sits between the two: a strongly held conviction that remains open to revision when the evidence shifts.

Most of our trouble comes from collapsing the distinction, treating our faith as fact and our beliefs as facts, and then defending all of it as though reality were at stake. For communicators, a great deal of what a company and its stakeholders solemnly call “the facts” are beliefs in disguise, and beliefs, unlike facts, can be worked.

The ‘lie’ inside Belief

The line that drew the loudest murmur of recognition was the simplest. Look, Eyal said, at the word itself. There is a lie sitting in the middle of belief. A belief does not have to be true to be useful. It only has to work.

A purpose statement, a positioning line, a set of values bolted to an office wall: none of these is a fact in any meaningful sense. They are instruments, and the only honest question to ask of them is whether they move people in the direction the business needs.

Nir reached for Amazon as his worked example (based on an ex-Amazonian sitting in the audience). One of the company’s founding internal beliefs is that it is always Day One. This is obviously nonsense. Amazon is a vast, mature, institutionally fortified machine, and it has not been on its first day for decades. As a tool it is close to priceless, because a company that genuinely believes it is still scrappy and still at risk behaves like one, and a company that believes it has arrived may find its energy sapped away.

Reputation is a belief held by other people

Start with what a reputation is. If beliefs filter what an audience notices, feels and does, then a brand reputation is simply the belief a stakeholder already carries about your organisation. That is the asset communicators manage, and it explains why disclosure on its own so rarely repairs trust.

We tend to behave as though the facts, once released, will speak for themselves. They never do. The facts arrive, and the pre-existing belief decides what they are permitted to mean. Nir put a number on the gap between what reaches us and what we register: the brain takes in something like 11 million bits of information a second and consciously attends to roughly 50 of them. We see more of what we already believe and screen out the rest. Pour more facts, more proof points and more content into that system and you change very little, because the belief doing the filtering has not moved. The work is on the filter, not the feed.

The real test is what a brand stands for when it costs something

If a belief is a tool, then the test of it is not whether you can recite it on a good day. The test is whether it holds when it costs you something. Nir’s point aimed at organisations is that companies go wrong at the exact moment they abandon their beliefs. Everyone is a person of conviction when its easy to do so. The interesting question is what gets quietly dropped when a situation becomes challenging: when the regulator calls, when the share price moves the wrong way, when the easy thing and the stated value point in opposite directions.

The gap between the belief a company professes in its annual report and the belief it reveals under pressure is, more or less, the entire territory in which reputations are won and lost. A value that is jettisoned the moment it becomes expensive was never a belief.

We treat a crisis as a contest over facts, a scramble to establish what happened and in what order. It is that, and it is a belief test first. The most common failure I see is the one Nir describes, as learned helplessness: the retreat into a defensive crouch, the decision to go quiet, say nothing and wait for it to pass. It feels protective. It is usually corrosive, because silence is read, fairly or not, as the absence of any belief worth defending.

Why a vague brand narrative is no narrative at all

There is a fashionable counsel of caution abroad at the moment which holds that, in an age of permanent uncertainty, the smart move is to commit to as little as possible. Hold every position lightly. Keep the narrative vague enough that nothing can later be held against you. I have argued before on White Wiki that uncertainty is exactly the condition under which communicators should be saying more, not less, and Beyond Belief sharpens the point.

A belief held so loosely that it survives any circumstance is not flexibility. It is the absence of conviction wearing flexibility’s clothes, and audiences can smell it from a great distance. The organisations that come through a hard year with their reputation intact are not the ones that believed nothing and therefore risked nothing. They are the ones that chose their beliefs deliberately, in advance, knowing what those beliefs might one day cost, and then paid the bill when it came due.

The beliefs worth holding are the ones that make you sweat

Which brings me back to the taxi. Eyal is fond of the line that everything worth having sits on the other side of discomfort, and that the trick is learning to manage the discomfort rather than wait for it to lift. I cannot pretend that an hour in a sweatbox on snail-slow London roads taught me anything profound about the human condition. But it is a serviceable reminder that the beliefs worth holding are precisely the ones that occasionally make you sweat. Perhaps the brands that survive the test of time do the same.

If you found this article thought-provoking, then you can buy Beyond Belief here.

Having now watched the events unfold at the other big tech conferences of the month, an uncomfortable question has been weighing on my mind… If June’s tech agenda was so focused on trust and AI, why was our discipline of strategic communications so hard to find on the stages?

How the tech conferences argued our subject

In London, the dominant human problem was trust. A friend working the events had been dropping the session summaries into a WhatsApp group, hundreds of them across London Tech Week and the London AI Summit, and trust appeared in 154 of the 352, more than any other theme. Around it clustered regulation, governance, the future of work, and the sovereignty question I wrote about last time.

And at SXSW London, a celebrity haunt as much as a tech festival, with Michelle Obama recording her podcast, the tech names on stage included Victor Riparbelli of Synthesia, whose business is synthetic video, and Rose Wang of Bluesky, whose proposition turns on trust and moderation. You can’t help but feel there is something subtly instructive about debating who the public can believe at a festival that is hosting synthetic-video pioneers and a former First Lady. Strip the lanyards off, and you are left with the management of public belief at scale - PR.

At VivaTech in Paris, artificial intelligence had stopped being a topic on the agenda and become the grammar of the whole event. The through-line the organisers kept returning to was sovereignty. A French agency even produced a tracker of the speaker list, which framed the event as a contest for European technological independence against the United States and China. Numerous ministers and regulators were in attendance, a sure sign that Europe is in the final, decisive stretch of writing the rules for how AI meets the public.

The questions on every stage - who is believed, who is trusted, who gets to tell the story - are communications questions. We have spent a century developing the vocabulary for them, but struggle to position ourselves.

We were not holding the microphones

With all those summaries to hand, I counted who was actually standing in the limelight.

Just 2 of 865 speakers at London’s two big AI events came from communications, public relations or corporate affairs. Not one of them was a chief communications officer.

Founders and chief executives took 287 of those 865 seats, technologists another 285. The vocabulary of communications appears nowhere across the 352 summaries. The word reputation itself surfaces twelve times.

Paris repeated the lesson in an even more pointed form. The independent tracker I mentioned earlier sorted VivaTech’s speakers and delegates by function, but did not trouble to give communications a category of its own. It folded the discipline into marketing, and even the combined pair came to just 4% of the room.

And going back to the two occasions in the London transcripts where communications did surface, it appeared first as a risk (”PR or legal issues”) and second as a line on a shopping list, a requirement to be procured alongside market access. We are filed, by the people building these systems, as a cost to be managed and a service to be bought, rather than as the discipline that owns the very problem they spent the month worrying about.

The room we were actually in

During the thick of tech conferences, the AI for PR Conference ran in London. It had a strong line-up: President of the CIPR, Farzana Baduel. Chief Executive of the PRCA, Sarah Waddington CBE. Carolyn Esser, Chief Corporate Affairs Officer at Darktrace, which is to say a communications leader at a listed cybersecurity firm, exactly the sort of person the London stages never thought to ask. Alongside them, much of the measurement community, the people now writing the rules for how a brand stays visible when an AI is the one doing the answering.

I was exploring Norfolk with the family, so unfortunately missed the occasion - a slight irony given what I’m writing about today - but the programme is exactly what our industry needs to care about. Reputation, crisis, governance, the integrity of the function in an automated world.

Yet we’re still talking in a bubble, and in the next 12 months we need to burst through to appear on those stages.

Three uncomfortable questions

It’s disappointing that, as someone who invested time and money in a CIPR-accredited PR degree, has since spent 15 years in the industry, and takes education as a professional responsibility, I feel the very industry I’ve bet on is in a perilous position when it comes to AI.

These are the questions every communications practitioner should be asking themselves:

How do we stop being the people who organise the stage and start being the ones the audience came to hear? There is a huge importance in coming together as an industry to share and learn from one another, but businesses are embedding AI into their operating models, and the integration and rollout of AI are largely communications challenges to navigate.

Have we made ourselves legible to the people setting the rules? Our industry is hugely diverse: representing all sectors and specialisms, with businesses ranging from freelancers to multinational giants. Perhaps our variety as a function creates buyer confusion around the concept of communications, meaning board-level appointments aren’t universally understood?

Are we spending our scarce attention on the right emergency? The tooling, the workflows, the very absorbing question of our own productivity are all real. But the categories that will govern how AI meets the public are being written this year and next, while they are still loose enough to shape, by people who do not think of trust as ours to hold. That, and not the content pipeline, is the thing with a closing window.

So what room are we building?

In the London Tech Week piece I ended on ownership: someone always owns the thing, and someone always tells the story of who owns it, and your only real choice is whether that someone is you. The same logic holds one level up, for the profession itself.

There are always two rooms. The one where we explain ourselves to each other, and the one where everyone else decides what we are for. We have become very good at filling the first. We cannot keep leaving the second half empty and wondering why nobody saved us a seat.

In the next 12 months, it would be good to see the role of strategic communications elevated industrywide, and to see more industry colleagues on those tech stages.

The story of London Tech Week 2026 rested on two themes: capital and sovereignty. Let’s start with the capital…

London Tech Week launched with a fanfare of announcements.

• £2 billion: AMD’s commitment to the UK over the next five years.

• £1.7 billion: Nebius committed to building out AI capacity in the UK.

• £400 million: Sir Keir Starmer used his keynote to set out a national compute strategy, including to buy specialist AI compute, alongside a wider AI hardware plan worth around £1.1 billion across chips, compute and skills announced by the Technology Secretary.

• £45 million: Lord Vallance announced the Engineering Biology Value Chain Growth Fund, on top of larger sums for research and scale-up infrastructure.

• £12 million: Mayor of London, Sadiq Khan announced investment into helping London’s small firms adopt AI, and published a City Hall report warning that requests to connect new data centres to the capital’s grid already run to roughly ten times the power those centres draw today.

An accountant for UK PLC may say these investments build the resilience necessary to cement Blighty in Europe’s so-called “decisive decade”.

Then on sovereignty, this was a full-on informal state occasion. The Prince of Wales joined a panel on using data to spot the risk of homelessness before it arrives, and his Homewards initiative launched a Homelessness Data Lab. The Mayor of London launched a programme of his own. The Prime Minister opened London Tech Week, as he did last year. The Science Minister attended, the Minister for AI, the list was quite endless… even the Director of GCHQ – an eagle-eyed observer would spot discreet security with earpieces.

So the two forces, sovereignty and capital, and what they mean for us humble communications practitioners.

The first force: sovereignty

Digital sovereignty is a grand term for a series of basic questions. Who owns the compute, the energy, the data centres and the data that artificial intelligence runs on? Who sets the terms for everyone else who uses them?

Sovereignty is multidimensional, and every country defines it to suit its own geopolitics, economy and security. The United States, the European Union and China hold three competing philosophies, written into the US CLOUD Act, the GDPR and China’s data security regime, and most other nations arrange themselves somewhere along that spectrum.

The sharpest version on stage came from Judith Dada of Visionaries Club, on a panel with the AI minister Kanishka Narayan, HPE’s Matt Harris and George Osborne, the former UK chancellor who now runs OpenAI’s work with governments. Dada argued that the prize is owning the bottlenecks that confer leverage, and noted that Europe holds a sliver of the world’s AI compute capacity, compared with roughly four-fifths in the US. Osborne made the realist’s case, that trying to own every layer of the stack is a fool’s errand, and that a country should aim to be indispensable rather than self-sufficient.

The consensus, once you strip away the rhetoric, is quietly pragmatic. No country is going to build the entire stack itself. Complete independence is neither affordable nor desirable, and a home-grown system would struggle to keep pace with the frontier. The serious goal is leverage and flexibility: control over the parts that matter, and the freedom to move when the politics or the economics shift. In practice, sovereignty is a negotiated set of options. The fortress was always a fantasy.

And it is not only a question of hardware. A sovereign capability is worth nothing without a workforce that can use it and a public that trusts it. Skills, adoption and public confidence sit alongside compute and data as conditions of the same independence. A nation, or a company, that owns the infrastructure but cannot carry its people with it has bought a tool it cannot lift.

The second force: capital

The second force was quieter on the main stages and more honest in the numbers. Britain is brilliant at building companies and less good at keeping them. The figures in this year’s Tech Nation report are strong. UK AI start-ups raised £8.2 billion in the first half of 2026, the country took close to half of all European tech investment, and the wider tech sector is now valued at around £1.2 trillion.

Read the same report a page later and the catch appears. Roughly fifty-seven pence of every pound made on an AI exit flows back to the US, with only nine pence staying here, and about one in two venture dollars invested in British AI arrives from across the Atlantic. The British Business Bank’s chair, Stephen Welton, spoke about scaling deep tech, which the UK has learned to do well. The harder problem is ownership at the finish line.

The example everyone knows, and few said aloud, is DeepMind. A London company, co-founded by Demis Hassabis, who by his own account wanted to keep it in London, sold to Google in 2014 for a reported £400 million because it needed the capital, the data and the computing power to stay at the frontier. The capital and the bottlenecks were in California, so in time the ownership, and the value, followed them there. The company still sits in King’s Cross; the cheques are signed in Mountain View.

That is the moment the two forces become one.

Whoever owns the compute and writes the cheque tends to end up owning the company, and with it the story of where the value was made. Mayor of London, Sadiq Khan was candid about the gravity… London has reclaimed its place as Europe’s leading tech hub and yet talent is mobile and can leave for Lisbon, Milan or Silicon Valley without much ceremony. AMD’s £2 billion and the growing London offices of OpenAI and Anthropic are welcome, but we’re currently a river that flows west.

The floor told the truth the stages skipped

Like last year’s London Tech Week, the expo floor – which I’ll certainly enjoy not experiencing for another year(!) – was full of conversations that coloured the corporate and state announcements. A good share of the stands belonged to other nations, each pitching itself as the place to incorporate, relocate and start again, their branding aimed at arrivals more than at home-grown firms with traction looking to expand.

The stage said build here. The floor said come and start over there.

Sovereignty at street level is a daily decision taken by founders and capital that can always choose somewhere else, and capital, as we have seen, usually has a view on the matter.

Regulation as a sovereign asset

Regulation itself can be a sovereign asset. British law firms were held up as leading adopters because the UK lets them experiment where the US does not. The medicines regulator may become the first in the world to license a gene therapy whose trial data showed it slowing Huntington’s disease progression by three quarters, data the American regulator has so far declined to accept. Owning your own rules, and using them with nerve, is a form of sovereignty that needs no hyperscaler’s budget. Although I’m not sure I want to be in a country known for rules over innovation – Keir Starmer’s opening speech didn’t help with that.

Sovereignty is a communications challenge

Everything is about communications, especially at a London Tech Week when capital and sovereignty are locking horns with one another. The questions a state now asks about compute and capital are the questions I hear from businesses every week. Here are five communications aspects to consider:

1. Every company needs a position on the sovereignty debate and decide if that needs to be public. For a growing number of companies, the debate is unavoidably public; if you run critical infrastructure, hold sensitive data, sell to government, operate a UK arm of a US parent, or depend on a single foreign platform, your customers, regulators and staff increasingly expect a view on where your technology comes from and who controls it – you must be prepared. This has already moved from reputation into disclosure. AI governance now appears in proxy statements and annual reports, investors expect boards to explain how AI is overseen, we even see this in those lengthy and tedious agency RFP processes. AI risk has joined cyber and supply chain as a standard line in corporate reporting. Of course, silence is a position as well, but you’re only really delaying the inevitable.

2. Know where your data lives, and who can compel access to it. The whole industry is fixated on which model to use. Far fewer leaders can say which jurisdiction holds their data, which sub-processors touch it, whether their providers train on it, and how much contractual freedom they have to move it. The lesson from the sovereignty debate is that location has stopped being the point. A data centre in Paris run by an American operator can still fall under American law. Control is the asset, and geography no longer guarantees it. For a communicator, that is a governance question with sharp reputational and disclosure edges, and the one most likely to surface in a regulator’s letter, an investigative piece, or the answer an AI system gives about your company.

3. Do not promise an independence you cannot deliver. The honest national lesson, that no one builds the whole stack alone, holds for companies too. Trying to engineer your own sovereign capability is expensive, hard to maintain and destined to trail the frontier. The credible story is flexibility: a setup built on solid foundations, with the optionality to change supplier, jurisdiction or model when you need to. Retire the sovereignty theatre and describe the real posture, which is pragmatic dependence held under firm control.

4. Remember that the machine is now a stakeholder. Your reputation runs partly on infrastructure you do not own, and the most important reader of your words may be generative AI. The answers across the prompt lines of Claude, ChatGPT, Perplexity, and so on, describe you to buyers, journalists, investors and regulators. They read what you publish literally, and they repeat their account of you to everyone who asks. Clarity and verifiable evidence are now the price of being described correctly.

5. Make your employees and your broader stakeholders part of the story from the outset. The sovereignty debate keeps returning to skills, adoption and confidence, and so should your communications. The internal narrative, how your people are trained and brought along, and the external one, why the public should trust what you are building, carry as much weight as any infrastructure announcement. Capability that no one understands or trusts is not capability at all.

So what did I learn from London Tech Week 2026?

AI has turned infrastructure into politics, and politics into a capital requirement, and communications now sits where the two meet. Strip the week back and the discipline for any company is three questions. Where do you stand on who controls the technology you depend on, and is that position ready to be public? Where does your data live, and who can compel access to it? And who tells your story when you are not in the room, whether that reader is a regulator, a journalist or a machine?

Someone always owns the thing, and someone always tells the story of who owns it. Your only choice is whether that someone is you.

Each month, the Press Gazette ranks the 50 largest English-language news websites, and I fear for my reporter friends trying to survive in these times.

The mastheads our parents trusted, and our clients still pay us to monitor and appear in, are still struggling to find a business model that rewards perspectives and news.

The Reuters Institute‘s annual Digital News Report explains the migration in prose only slightly less funereal than my own: trust in traditional publications eroding, subscriptions stagnating, audiences drifting from the news brands to individual reporters. It’s these days that the columnist, the commentator, the named human with a discernible point of view, matters to readership.

One media brand is going against the grain: Substack. It’s the media brand that boasts growing traffic levels despite having no copy of its own, no large newsroom of reporters, and existing mainly to email other people’s writing to the people who asked for it. Its traffic has climbed from roughly 95 million visits at the end of 2024 to north of 169 million, very nearly doubling at the precise moment the rest of the industry was learning to do more with less.

This may explain why I’m frequently asked by a brand or an executive, “Should we be on Substack? Should I?”

It is a good question, and here is the answer.

I should declare an interest first: this very publication began life as a LinkedIn newsletter called White Space, and I picked it up, subscriber list and all (yes, there is a method…), and carried it over here. So clearly, I have voted with my feet (or my fingers?).

You’re already on LinkedIn

For most of the past decade, the answer to “where should our executive build a profile?” required no thought at all. LinkedIn. Obviously LinkedIn. It is where the professional audience gathers, the networking is frictionless, and the format forgives a great deal.

Then the foundations changed.

Over the last 12 months, LinkedIn rolled out a recommendation engine its engineers call 360Brew - a large language model, reportedly some 150 billion parameters, now deciding what the feed shows you. It reads your actual words, infers what each individual reader privately cares about, and renders a verdict on your relevance. The consequences have been well documented.

Richard van der Blom’s Algorithm Insights study, drawn from 1.8 million posts, found views down 50%, engagement off 25%, and follower growth slowed by 59%. Company pages fared worst of all: organic reach down somewhere between 60 and 66% since 2024, with the average corporate page now reaching around 1.6% of the followers it worked so hard to acquire. You may recall this is a transition Facebook went through around 2016.

Even LinkedIn’s own stars feel the chill. Simon Sinek, Adam Grant and Steven Bartlett command audiences in the millions there, and all three now work a feed that rations attention more tightly than at any point before. However, keep in mind that it’s a personal profile now that earns several times the reach of a company page sharing the very same words.

Now clearly, investing time in LinkedIn is still important as it remains the largest professional gathering on a social network with the backing of Microsoft. However, building a profile on LinkedIn is like investing in a house on leased land and everything has the potential to quickly change. Keeping this homeowner analogy going, there is nothing quite like owning your own place.

What to consider on Substack

Earlier this year I attended an event with Emma Rowley, Substack’s Head of News and Culture UK, to explore what it takes to be successful on Substack.

The room was full, and the air was distinctly hot with communications people - agency and in-house - every one of them circling the very question I now field. You could watch the penny drop as the room understood that the platform has remarkably little interest in the organisation. It wants a person.

For most executives and brands, Substack won’t be right. Mostly as they’re unwilling to do the one thing Substack rewards, which is an authentic human voice that offers a perspective worth considering. If your internal approvals process includes a dozen people and multiple rounds, we all know that the outcome is a piece of polite airline journalism that doesn’t require thought.

BUT, if you are bold, brave, and belligerent (no, really), Substack could work. Have something to say, be credible with it, and let the institutional glaze of responsibility take a slight breather - be willing to connect with people and exchange ideas in a public forum.

Owning your audience has quietly become one of the very few moves in executive communications where the wind is at your back rather than in your face. The inbox, it turns out, is about the last piece of land you actually own. I would think hard about who, in your organisation, deserves to build on it.

LinkedIn vs Substack, the benchmark

Let’s pit both platforms against each other, but they’re very much bedfellows in reality.

On search, the gap is structural. A LinkedIn post lives behind a login wall and is, to a search engine, very nearly invisible; only LinkedIn’s long-form Articles are reliably crawled and there is a small chance Gen AI responses will surface them. Every Substack post sits on the open web, on a domain you can make your own, indexed and accumulating the topical authority that search rewards.

That second clause is the one most comms teams have yet to price in. When a journalist, an analyst or a prospective client asks ChatGPT or Perplexity “who is the leading voice on X?”, the model answers from what it can retrieve, and it retrieves from the open, indexable, persistent web far more readily than from a walled feed.

The sharpest operators are now building what the Generative Engine Optimisation (GEO) crowd call triangulated authority: a personal site, a LinkedIn presence and a Substack, cross-linked so that an executive resolves, in a machine’s eyes, into a single verifiable expert. An owned, indexable body of work has become a reputational asset in its own right, and a login-walled feed is the wrong place to keep one.

A practical caution while we are here. LinkedIn now suppresses posts carrying external links by something like 40% to 60%, so the old “here’s my newsletter, do click through” play actively backfires. Lead with the idea natively, hold the link back, and route readers across with rather more subtlety than the job used to require.

So… should you, or shouldn’t you?

If the macro case has tempted you, the decision still comes down to a short and unsentimental set of questions. I would not proceed unless your executive clears most of them.

Have they got analysis, or only announcements? Readers subscribe for thinking that sharpens their judgement - a framework, a contrarian-but-defensible view, a lesson from the actual coalface of running something. A press release in a nicer typeface won’t work.

Will they write it, and keep writing it? Cadence is the single best predictor of whether one of these survives. The internet is an ossuary of executive newsletters that launched to fanfare and fell silent by week seven, and a dead Substack speaks louder than no Substack.

Can you let it run free of the sales function? The publications that earn the regard of journalists and analysts read as editorial. The instant yours smells of the pipeline, the credibility that justified the whole exercise evaporates, and you are left with an expensive brochure nobody opens.

Can your sign-off keep pace? Regular publishing needs a review process robust enough to protect the business and quick enough to clear a post within the news cycle it belongs to. Three approvals and a fortnight’s delay will strangle a publication before it draws breath.

Can you live with the comments? A public platform hands your critics a permanent venue, and someone will need to tend it. For an executive with active detractors that is a genuine consideration, though a manageable one if you have planned for it rather than discovered it.

Do you have the time for Substack Notes? It’s widely reported that Substack Notes is the best growth engine for subscribers and views, but it’s akin to managing an old-school Twitter account. You need daily publishing, comment fielding, and open interaction with other authors.

So, what do you think? Ready for Substack?

There is a particular flavour of frustration, familiar to anyone who has ever tried to get a legitimate, properly sourced, conflict-of-interest edit actioned on Wikipedia, that I have come to think of as the encyclopedia’s quiet tell. You do everything by the book. You declare your interest. You assemble your citations like a barrister preparing a brief, reaching for the independent secondary source rather than the convenient primary one. You place the little edit-request template at the foot of the talk page, append your courteous request, sign it with the requisite four tildes, and then you wait. And wait.

Unless you have the rare convenience of an active editor on the page, most COI requests join a centralised queue, patrolled by a small but dedicated band of volunteers. The oldest open requests in that queue tend to hover somewhere between three and four months old. Some have been seen wearing the dust of a year or more. Somewhere around week six, you begin to suspect that the encyclopedia which now functions as the closest thing humanity has to a shared factual substrate is being maintained by roughly the number of people it takes to staff a mid-sized garden centre.

That suspicion is the puzzle this article sets out to solve.

On the one hand, Wikipedia has never mattered more, for reasons I will come to that have everything to do with artificial intelligence and almost nothing to do with its readers. On the other hand, getting a careful and entirely legitimate change made to it has never felt slower. I went in believing the obvious explanation, the one most people believe: that Wikipedia is simply running out of editors, fewer hands at the tiller every year. That turns out to be wrong. What is actually happening is stranger, and rather more alarming, than simple decline. Here are four facts, with four graphs. Take them in order.

PART ONE: THE EVIDENCE

Shift one. It is not, in fact, dying

The popular narrative has a comfortable shape. Wikipedia peaked in its idealistic youth and has been bleeding contributors ever since. There is real history beneath this. The number of active editors on English Wikipedia, defined with bureaucratic precision as registered, non-bot users making five or more edits in a given month, did indeed peak in 2007 at somewhere north of 51,000, and then entered what the researcher Aaron Halfaker memorably christened the decline phase. His widely cited 2013 study traced the cause not to apathy but to the encyclopedia’s own immune system: as Wikipedia armoured itself against vandalism with bots, automated reverts and an ever-thickening hedge of policy, it became markedly less hospitable to newcomers, whose first tentative edits were now liable to be reverted by an efficient and impersonal machine. The newcomers, unsurprisingly, declined to return.

But here the data does something the narrative does not. The decline did not continue to zero. It bottomed out around 2013 to 2014, at roughly 31,000, and then it stabilised. The current figure, the one sitting on the encyclopedia’s own statistics page as I write, is around 37,000 active editors a month. Below the giddy high-water mark of 2007, yes. But comfortably above the trough, and broadly flat for more than a decade.

So the obituaries were premature. English Wikipedia is not dying. The patient is, if anything, eerily stable. Which raises a far more interesting question. If the workforce has held steady at around 37,000 for ten years, why does it feel, to anyone actually trying to get something done, as though there is nobody home?

Shift two. But the house has doubled

The answer is that we have all been watching the wrong number. The headcount is not the problem. The ratio is the problem.

+62% Growth in English Wikipedia articles since 2013, from 4.4 million to over 7.1 million, while the active editor base has stayed broadly flat. The same staff now maintains half as much house again.

In 2013, when the decline narrative was being written, English Wikipedia held about 4.4 million articles. Today it holds over 7.1 million. The encyclopedia has grown by more than half again, nudging towards a doubling, while the body of people maintaining it has not grown at all. Every one of those new articles must be watched, sourced, defended against vandalism and decay, updated as the world turns, and dragged into line with two decades of accreted policy. The house has gained wings, attics and a great many leaking gutters, and the same modest staff is expected to keep all of it weatherproof.

It is worse than even that ratio suggests, because the labour is grotesquely concentrated. The cliche, tiresome but essentially true, is that something close to 1% of editors produce the overwhelming majority of the content. The genuinely heavy lifting falls to the very active cohort, those making a hundred or more edits a month, who number only around 3,000 - 3,500 souls in any given month. When you submit an edit request, you are not appealing to a crowd of 37,000. You are joining the in-tray of a few thousand volunteers, most of them doing this in the evenings, for free, and who were, even a decade ago, described by the Foundation’s own people as feeling beleaguered and overworked.

Shift three. And the officers have all but gone

If the editors are merely overstretched, the administrators are something closer to an endangered species, and this is where the practitioner’s frustration finds its true source. Administrators are the editors entrusted with the heavier tools and, crucially, with the authority that often decides whether a contested edit request lives or dies. They are anointed through a process called Requests for Adminship, and the collapse of that process is one of the more astonishing graphs in the whole of internet governance.

The number of administrators today stands at around 820, a figure that has drifted downward for years and which flatters the reality, since a good share of those accounts are dormant. The process had become so notoriously gruelling, so much like running a gauntlet of strangers entitled to interrogate your every edit from a decade past, that qualified candidates simply stopped putting themselves forward. Matters had grown sufficiently dire that in 2024 the community did something it almost never does: it changed the constitution, introducing secret-ballot administrator elections expressly to spare candidates the public flaying of the old process. That is not the behaviour of a healthy institution casually iterating.

Shift four. While the readers are quietly intercepted

Wikipedia has never mattered more than it does today, and the reason has nothing to do with its readers. It is because of the machines. To a first approximation, every large language model in commercial use has been trained on Wikipedia, and the Wikimedia Foundation admits that its corpus is almost always the single largest source in those models’ diets. It is probably over-represented even relative to its size, because it is mirrored so promiscuously across the web that it ends up baked into the training data several times over. One researcher, reaching for a suitably undignified image, described the whole training corpus as a giant hairball, with Wikipedia threaded all the way through it.

It does not stop at training. Wikipedia underpins Google’s Knowledge Graph, populates the knowledge panels, feeds the AI Overviews and the chatbot answers, and serves as the primary credibility checkpoint against which a Generative AI model decides whether a claim about you, your client or your company is to be trusted. When an AI confidently describes a business to a user, it is very often Wikipedia doing the talking, ventriloquised through a model that will never cite it.

And there is the rub. The machines that depend on Wikipedia are simultaneously starving it.

−23% Fall in daily human visits to Wikipedia since early 2022, from roughly 165 million to under 128 million, with an 8% year-on-year drop in 2025 alone once disguised bot traffic was filtered out. We are all still reading Wikipedia. We have simply stopped visiting it.

The Foundation’s own analysis puts the cause plainly: search engines increasingly answer the question directly, in a tidy package, with no click required. When I interviewed Jimmy Wales at the end of 2025, he put the same figure to me himself, the 8% drop in human visits even as bot traffic climbed, and described the disproportionate hit on a charity’s server costs as AI firms crawl the site around the clock. He was, characteristically, more sanguine than alarmed. For a charity, he argued, lost page views matter less than lost attribution; the real sadness would be a world in which the public no longer knows the knowledge came from Wikipedia at all.

That is the optimistic reading. The pessimistic reading, which the graphs above quietly support, is that readers are not only an audience but a recruitment funnel. The casual visitor who fixes a typo today is the very active editor of 2030; the reader who sees the donation banner is the reason the servers stay on. An AI layer that consumes Wikipedia’s knowledge while intercepting its readers is, in the bracing word now used across the publishing industry, parasitic. It feeds on the host while cutting off the host’s supply of nutrients. The decade of flatness on current trends, is about to get a great deal less flat.

The snake and its tail

In March 2026 the community drew a line. After a Request for Comment that had failed in various forms before, English Wikipedia voted, by a margin of roughly forty to two, to prohibit the use of large language models to generate or rewrite article content outright. The exceptions were narrow and telling: you may use an AI to copyedit your own prose, provided you check it, and to rough out a translation. You may not use it to write the encyclopedia.

The reasoning was not technophobia. It was arithmetic and self-preservation. Generating a plausible, confident, entirely fabricated paragraph now takes a machine seconds. Detecting it, verifying it against sources and cleaning up after it takes a human volunteer hours. In a community where volunteer-hours are precisely the scarce resource, that asymmetry is fatal. Wales sketched the underlying mechanism to me in plainer terms than most policy documents manage: a model simply picks the most probable next word, so the most plausible thing is not the same thing as the most true thing, and what you get, eventually, is hallucination dressed as fact. He was unambiguous that the Foundation should never put AI-written text in front of readers without a human checking it first. The editors went further still, and named the deeper trap: the compounding loop in which hallucinated text slips into an article, is scraped into the next training run, and re-emerges from a future model as apparent fact, complete with a citation trail that loops back, eventually, to the lie. The serpent eating its own tail. The straw that broke the camel’s back, fittingly, was an autonomous agent that had taken to authoring articles on its own initiative.

PART TWO: THE CHALLENGE TO COMMUNICATIONS

So gather the four shifts together, because their sum is the answer to the puzzle I began with.

A flat workforce. Maintaining a doubled encyclopedia. With a collapsed officer class. And a reader funnel now being drained by the very machines that depend on the output.

Your conflict-of-interest edit request does not languish because Wikipedia is hostile to you, nor because you have done it wrong. It languishes because it is the downstream symptom of every one of those graphs at once. It sits near the bottom of a months-long queue tended by an over-stretched few who are also, as of this spring, fighting a war against machine-generated content. That is the whole mystery, solved.

For those of us in the reputation and intelligence trades, this rearranges the problem in three uncomfortable ways.

1. The stakes have inverted and risen. It used to be that an error or an unflattering line on a Wikipedia article cost you one webpage, read by whoever happened to land on it. That is no longer the unit of damage. Because Wikipedia is now the credibility checkpoint for the models, an outdated or wrong article propagates outward into AI Overviews, chatbot answers, voice assistants and knowledge panels, becoming the machine consensus about your client, repeated confidently and without attribution to a public that increasingly never visits the source. The article is no longer a page. It is a seed crystal for everything the machines will say about you next.

2. The legitimate route has narrowed exactly as the stakes have widened. This is the cruel scissors of the whole situation. The proper, disclosed path, the edit request made under the paid-contribution rules, depends entirely on the attention of experienced editors and administrators, which is the one resource every graph above says is shrinking in effective terms. The correct way to do this has never been slower. Meanwhile the improper shortcuts, the undisclosed account, the quiet direct edit, are more dangerous than they have ever been, because the community’s new vigilance against both conflicted editing and AI-generated text means anything that looks engineered is more likely to be caught, reverted and pinned to your client’s name in a place that the models will then dutifully learn from.

3. You cannot simply automate your way around the bottleneck. The obvious temptation, in 2026, is to have a model draft the polished article and submit it. That door has just been bolted. Generating or rewriting article content with an LLM is now against policy and actively policed, and while reviewers have sensibly agreed that prose merely reading like AI is not by itself a hanging offence, the burden of suspicion has shifted. The machine that made your job look easy has made the gatekeepers trust the gate less.

What actually works, then, is unglamorous and worth stating plainly. It is also, more or less, what Wales himself described to me as the approach that succeeds. Do not try to turn an article into a PR puff piece; that, he was frank, is simply not on the menu. Instead, point out where a published fact or the other half of a story has been left out, supply the reliable source, and let the evidence carry it. Arrive with the work finished: a clearly formatted, drop-in change, sourced to independent secondary material rather than the company’s own announcements, and obviously in service of a better encyclopedia rather than a shinier subject. And build a reputation.

On Wikipedia, as Wales put it to me, your pseudonym is your reputation, earned slowly through consistent and visible good behaviour; editors listen to the people they have learned to trust, and revert the ones they have not. The scarce resource is volunteer attention, so the winning move is to spend as little of it per request as possible. There is no growth hack for an under-resourced platform. There is only making the volunteers’ lives easier, and waiting your turn.

The authority of Wikipedia has never been higher. The number of hands holding it up has never been more precarious relative to the weight. Those two lines cannot keep travelling in opposite directions forever.

The lesson for anyone whose work touches Wikipedia is therefore not to treat its remaining editors as obstacles to be routed around, but as the overstretched custodians they actually are, of a thing far more important than the average comms brief allows. It is also, I think, the lesson Wales kept returning to in our conversation: trust is not a static asset but a thing built, slowly, through purpose, evidence and openness, and just as easily lost. The least we owe the people still doing that building, before we ask the machines anything else, is to remember whose answers we are really receiving.

I have spent the time since wading through the sessions and the largely tedious industry reaction. The loud quarrel of the past two weeks, the one pitting evangelists who pronounce Search Engine Optimisation (SEO) dead against purists who dismiss Generative Engine Optimisation (GEO) as a manufactured buzzword sold to the credulous, is the wrong one. Both camps are still arguing about discoverability: how a brand gets found. I/O 2026 was not, at bottom, about how you get found. It was about who now owns the act of discovering and understanding on the public’s behalf, and what that ownership does to the economics and the surveillance of reputation. That is a different conversation, and for our profession a far more uncomfortable one.

Let me start with the facts, because they are better than the rhetoric.

What actually happened

AI Overviews now reaches over 2.5 billion monthly users. AI Mode has crossed one billion monthly users in roughly twelve months, having grown something like tenfold since the previous autumn. Google’s own framing is the part worth dwelling on: queries are more than doubling every quarter, and total search volume reached an all-time high last quarter. People who use the AI features are not searching less. They are searching more, because the friction of composing the perfect keyword string has been abolished.

Around those numbers Google announced four things that matter to anyone who manages a reputation for a living.

1. The search box itself has been rebuilt for the first time in over twenty-five years. It is now multimodal by default, accepting text, images, files, video and even open Chrome tabs, and it reasons across all of them at once. AI Overviews and AI Mode have been merged into a single flow, so a user drifts from a results page into a conversational back-and-forth without ever choosing to. AI Mode has not been made the formal default, but the behaviour is now so close to default that the distinction is largely academic.

2. The model underneath it all is Gemini 3.5 Flash, now the global default in AI Mode, fast and cheap enough to run agentic tasks at the scale of the entire search index.

   

3. Google introduced background information agents. These are autonomous systems that sit inside Search and monitor a topic continuously, looking across blogs, news, social posts and live data, then surfacing changes without anyone having to run a query. I will return to this, because it is the most underrated announcement of the week.

4. Google is extending its SynthID watermarking and C2PA Content Credentials into Search and Chrome. SynthID has now marked more than a hundred billion AI-generated images and videos, and partners including OpenAI and ElevenLabs are adopting it. The provenance check is live in the Gemini app today and is rolling out to Search and Chrome in the coming months, which means the ability to ask of an image or clip, “was this made by AI,” is on its way to becoming a one-tap consumer reflex. For anyone defending an executive against synthetic media, that is good news.

So far, so consistent with what you have already read.

The number that tells the real story

In the same quarter that Google was celebrating record search volume, Alphabet’s Q1 2026 results showed Google Search revenue rising 19 per cent year on year to around $60.4 billion. In the very same results, Google Network revenue, which is the open web of AdSense and Ad Manager publishers, the sites that Google’s index was historically built to send people to, fell by 4 per cent to roughly $6.97 billion.

Read those two figures together and the strategic picture snaps into focus. Google is monetising synthesis on its own surfaces at a remarkable rate, while the open web from which it draws that synthesis is being starved by the very machine it used to feed. The traffic that funded the journalism, the reviews, the forums and the explainer content the models depend upon is declining even as the value extracted from that content rises. The single synthesised answer has become the destination, and when Generative UI can spin up a bespoke calculator, comparison table or dashboard inside the results page using the Antigravity engine, the user has no remaining reason to click through to your site at all.

This is why I find the SEO-versus-GEO row so beside the point. Both sides assume the game is still about earning a visit. The visit is the thing that is being eliminated. The optimisers are busy polishing the brass on a door that no longer opens. The question for a communications leader is no longer how to win the click. It is how to shape a judgment that is formed, delivered and acted upon entirely inside a system you cannot see into, cannot query and cannot appeal.

Why this is a category error for communications

There is a comforting consensus circulating among practitioners that AI visibility is “still eighty per cent good, fundamental SEO,” on the logic that large language models retrieve from the same authoritative indexes that traditional ranking rewards. As a technical observation about retrieval mechanics, that is true enough, and worth knowing. As a strategic posture for reputation management, it is a trap, because it flatters us into optimising outputs long after the leverage has migrated to inputs.

The deterministic web we grew up managing rewarded a recognisable set of outputs: a page that ranked, a link that was earned, a placement that could be screenshotted and reported. The probabilistic web rewards something stranger. A model does not care a great deal about your backlink profile. It cares whether your brand name is consistently associated with a particular claim across many independent and trusted nodes at once. It is, in effect, taking a weighted average of what the internet believes about you and serving that average back as fact, in the confident, citation-light register of a well-briefed colleague who has not quite done the reading.

That changes the unit of work. Page-one ranking is a legacy metric. The metric that now matters is share of citation: whether the model extracts, trusts and names you when it answers a question in your category. And share of citation is not won by volume. It is won by consistency and by source quality. A single, well-placed narrative in a tier-one title, a heavily cited industry report or an authoritative forum thread now outweighs fifty earnest posts on your own newsroom that nobody else troubles to corroborate. The strategy shifts from broadcasting to seeding: from producing content for audiences to curating the corpus the machine reads on the audience’s behalf.

The announcement nobody is frightened of yet

For as long as I have worked in intelligence, monitoring has been expensive and therefore rationed. A short seller, an activist investor, a regulator or a competitor who wanted to watch your every move had to commission people or platforms to do it. I/O 2026 democratised that capability and made it continuous and free. Anyone can now instruct a background agent to watch your sector under precise, nuanced parameters, to track every change, synthesise it and report back, around the clock, at effectively no cost.

The implication for crisis communications is structural. The weekly media-clippings model, retrospective by design, assumes that you and your adversaries watch at roughly the same cadence. That assumption is now false. Your critics are running always-on instruments while a good many comms functions are still reading yesterday’s coverage over this morning’s coffee. Worse, a misleading claim that takes hold in the sources a model trusts does not merely circulate; it risks being baked into the synthesised answer that two and a half billion people receive, repeated with machine confidence. The job is no longer to respond to a narrative. It is to detect and intercept one before it sets.

What I would actually do

If you accept that reputation is now mediated by a model before any human reaches your front door, the work reorganises itself around two verbs that have nothing to do with traffic: shaping the inputs, and shortening the time to detection.

On inputs, audit your entity footprint rather than your keywords. Test directly how Gemini, ChatGPT and Perplexity describe your company, your leadership and your contested issues, and find the gaps and the competitors being cited in your stead. Then build your owned assets, the newsrooms, investor pages and executive biographies, for extraction rather than persuasion: unambiguous, structurally clean, factually declarative, the kind of material a model can chunk and verify rather than skim and hallucinate around. And concentrate your earned-media effort on the narrow set of high-authority nodes that genuinely anchor a model’s answers, on the understanding that corroboration across trusted sources now beats volume on your own channels.

The bottom line

Google Search has not died, it has evolved and changed the web. The era of ten blue links offered a clean and frankly honest transaction: you optimised a page, the user clicked, and you controlled the room they walked into. You knew where you stood, even when you stood low.

In the agentic era that I/O 2026 confirmed rather than launched, that transaction is gone. Google is no longer the concierge in the lobby, pointing politely towards your door. Google is the host, standing on the threshold and summarising your entire story to the guest before they have decided whether to come in. The accounts tell us the host is keeping more of the takings while the suppliers who stock the kitchen are likely receiving less.

So stop managing the links. Start managing the story that feeds the machine, and watch that machine at least as closely as your critics already do. That is the real work the keynote handed us, and it began the moment the applause died down.

I used to find that question flattering. Now I find it unsettling. Because every time someone asks it, I’m not thinking about the person sitting across from me. I’m thinking about Michael White in 2008 - the version of me who was considering entering this industry, who had no idea what he was walking into, who thought communications was about the communications and not about the person doing the communicating.

So I decided to answer them properly. Not with platitudes. Not with motivational LinkedIn rhetoric. In total, these are the eight things I wish someone had told me when I was in their position. Things that have become obvious only because I’ve spent fifteen years in this industry and now find myself on the management team of a global consultancy, contributing to how this entire firm thinks about communications in an age of AI transformation.

If you’re considering joining this industry, or if you’ve just joined and you’re wondering whether you’ve made a terrible mistake(!), read on.

1. You’re only as good as your last piece of work

This is the hardest lesson and the most important. It is the lesson that keeps people honest. In communications, you can speak brilliantly about strategy in a conference room. You can command a meeting. You can have won a dozen awards for work that isn’t yours. And none of it matters the moment you deliver something mediocre, late, or half-baked.

This business doesn’t traffic in potential. It doesn’t care about your university degree, your internships, your connections - though all of those things help you get in the door. What matters is the email you sent yesterday, the deck you built last week, the brief you answered correctly when nobody else could see the angle. Your reputation in this industry is not a personal brand. It’s a work product. Get that wrong, and you’re starting from zero every time.

2. Understand people and their relationships

The industry will tell you it’s about media relations, project management, good writing skills, data analytics - and it is, sort of. But what it’s actually about is watching how people relate to each other and understanding why they do what they do.

You need to know what your client’s CFO wants from the CEO. You need to know why your team’s smartest strategist suddenly becomes quiet in meetings. You need to understand that the journalist who asks a brutal question in the first meeting might become one of your most valuable allies if you treat them with respect. You need to grasp that your colleague who seems invisible has actually watched three senior people fail at the same task and knows exactly where the trap is.

You cannot manage a narrative if you don’t understand the people telling it. You cannot manage a crisis if you don’t understand the psychology of the room trying to solve it.

3. Lean into people’s experience to innovate

You’ll arrive with fresh ideas. They won’t be new. Someone in the room has seen the version of this problem that you’re proposing to solve - maybe not exactly, but close enough.

The temptation is to ignore that. To present your idea as revolutionary. To treat experience like institutional resistance rather than institutional knowledge.

Don’t. Ask the person who’s been here longest what they’ve tried. Ask what failed and why. Then take that and push it further. That’s where innovation lives - not in the blank canvas, but in the willingness to stand on the shoulders of people who’ve already fallen trying.

At the same time, maintain a healthy scepticism of a broad consensus of an idea – if mankind’s ideas required to be agreed upon each time democratically, then our evolutionary desire for safety likely would have stopped progress. Henry Ford put it simply as this, “If I had asked my customers what they wanted, they would have said a faster horse.”

4. We’re all struggling with something. Don’t hide it.

A few years ago, I was not delivering my best work. I was forgetting things. I couldn’t process information the way I normally could. I’d sit in a meeting and feel present and completely absent at the same time.

I had just started taking medication for anxiety - connected to depression – and my brain felt like it was re-wiring itself. For a few weeks, I was in a fog.

Very few people knew. I certainly didn’t broadcast it. But looking back, I wish I had said something to the people who needed to know. Because they were probably wondering why I wasn’t myself. And I was probably wondering why they seemed frustrated with me. The gap between us grew because I thought strength meant silence.

It doesn’t. In a fast-paced environment where your intellectual performance is constantly measured, it is essential to recognize that all of us are humans. All of us are struggling with something. That might be anxiety. It might be grief. It might be a health crisis, a relationship falling apart, a mortgage that’s strangling you, a parent in decline, a secret that’s eating you alive.

You can’t build a functional team on the fiction that everyone is fine. You can’t build a client relationship on pretense. The moment someone on your team says, “I’m not okay,” you have a choice: you can pretend you didn’t hear, or you can actually listen.

Choose to listen.

5. This isn’t a job. It’s a way of life.

Strategic communications doesn’t have office hours. A crisis doesn’t wait for your holiday to end. A journalist doesn’t file at 9 a.m. and then call it a day. The news cycle, the social media pulse, the client emergency - none of it respects the boundary between your work and your life.

You need to go into this with your eyes open. This is a way of life. You’ll miss dinners. You’ll check your phone at midnight. You’ll be thinking about a client problem while you’re supposed to be present with your family. You’ll feel like you’re never quite off.

The flip side is this: if you’re the sort of person who loves the work, who genuinely enjoys solving problems under pressure, who gets energy from the chaos - there is almost no better place to be. But you have to know what you’re signing up for. You can’t do this half-heartedly. You can’t view it as a stepping stone and expect to be taken seriously.

6. Treat people as you want to be treated

This sounds like a platitude. It isn’t. It’s a survival strategy.

In communications, you will have power over people - over how they’re perceived, over their stories, over their reputation. You’ll have junior colleagues who want your approval. You’ll have vendors who need your business. You’ll have clients who depend on your competence. You’ll have journalists who might feature you or bury you.

The temptation to use that power loosely is constant. It’s easy to be brief with someone who isn’t useful to you. It’s easy to take credit for work you didn’t do. It’s easy to pretend to listen while you’re already thinking about the next meeting. It’s easy to be kind when someone can help you and distant when they can’t.

Don’t.

7. Rotten people spread amongst rotten people

This is the corollary. It matters because it’s true. Cultures don’t change slowly. They change through contagion.

If you work in an environment where corners are cut, where people take credit that isn’t theirs, where kindness is seen as weakness - you will become that version of yourself. Not because you’re weak. But because proximity to rot is corrosive. You’ll start small. You’ll justify it. You’ll convince yourself it’s necessary. And before you know it, you’re the person you said you’d never be.

During my career, I’ve seen this clash of cultures and values on the frontline. Sometimes it’s working with a client whose own internal way of working clashes with the values of your own team. In the past, it’s also been present when people join your consultancy directly from another, exhibiting behaviours that go against the grain.

This is why choosing where you work matters more than you think. This is why your first consultancy, your first team, your first boss - these are not trivial choices. They’re the choices that shape who you become.

8. Prioritize experience over pay, at least in the beginning

I can say this now because I’ve had the privilege of being paid well. But I came to it honestly. When I entered this industry, I didn’t negotiate for salary. I negotiated for exposure, for variety, for the chance to sit next to people who were better than me.

I was also in a privileged position. My family lived just outside London. I paid nominal rent. I didn’t have a mortgage. I wasn’t supporting anyone. I know that conditions aren’t the same for everyone, and that’s exactly why I support organizations like the Taylor Bennett Foundation - they exist because the path I took isn’t available to everyone, and it shouldn’t have to be.

But if you can afford to choose experience over salary in your early years, do it. Join the firm that’s working on the most interesting problems. Join the team where you’ll learn the most. Join the client where you’ll be pushed to your limits. The connections you make, the skills you develop, the reputation you build - these are currency in this industry. They’re worth infinitely more than a ten percent salary increase at the start of your career.

Because - and this is the thing nobody tells you - the pay comes. Not always fast, not always fairly, not always as much as you deserve. But it comes. It comes as a natural outcome of your experience, your reputation, and your relationships. The people who made different choices early, who prioritized salary over learning? Some of them are fine. Some of them are stuck. The ones who thrived were almost always the ones who understood that the first five years are an investment in themselves, not a paycheque.

The thing you actually need to understand

These eight points matter. But there’s something deeper that matters more.

This industry is transforming faster than any of us can keep up with. AI is rewriting what comms professionals actually do. The media landscape is fragmenting in real time. The relationship between journalists and PRs, between brands and audiences, between truth and narrative - it’s all in flux.

The candidates I interviewed who will thrive aren’t the ones with the cleverest prompts or the sharpest understanding of the latest platform. They’re the ones who understand that in an industry being rebuilt by machines, the only thing that can’t be automated is human judgment, human integrity, and human connection.

Come in with your eyes open. Work harder than anyone else. Be kind, even when it costs you. Build relationships like your career depends on it, because it does. And remember that the people who matter in this industry aren’t the ones shouting the loudest. They’re the ones who’ve kept their word, shown up for their team, and somehow managed to stay human while the whole world was going mad.

Good luck. And then you’re going to need something more than luck: you’re going to need judgment, resilience, and the kind of integrity that makes you someone worth working for.

The industry needs that. You can be that.

Christmas approaches, and I’m now managing a new demanding junior stakeholder, my beautiful son, who arrived on Sunday. With mum and baby doing well, my paternity leave has begun, and it is time to wave a thankful, if slightly exhausted, goodbye to 2025.

Before signing off for the year, however, I wanted to reflect on what I think 2026 may hold.

Intelligence to drive trust

Data-led insights continue to transform strategic communications into a function not just defined by media relationships but also elevate the roles of communications professionals to C-Suite decision makers; we protect and build a firm’s most important aspect, trust.

At the start of 2025, my focus began by naval gazing on the more polluted parts of the internet: misinformation and disinformation. I joined the firm Securitas for a talk in Stockholm, addressing the critical issues of narrative warfare and the communications landscape.

It was an apt narrative continuum following Mark Zuckerberg’s 7^th January 2025 announcement that Meta is abandoning the use of independent fact checkers on Facebook and Instagram, replacing them with community notes. You can read my independent thoughts on this subject by clicking here.

This thought leadership was reinforced at the end of the year by the privilege of interviewing Wikipedia Founder, Jimmy Wales, at a packed London event focusing on the fundamental concept of Trust. For a long time, I’ve been a supporter of the Wikimedia Foundation, and speaking about optimistically building trust beats focusing on the darker corners of the internet.

In between, I’ve been fortunate to speak and advise multiple companies facing challenges relating to their online profile. In particular, the ongoing impact of Generative AI. It’s clear that building trust in a zero-trust communications environment – accentuated by so-called AI slop and political polarisation – will continue to be a corporate challenge in 2026.

Zero-trust on social media is driving fragmentation

Social media continues to go through a catalyst of change. A data partner I work closely with, Pulsar, recently published their report ‘The Great Fragmentation: Navigating the Splintered Web’, showing how audiences, platforms, and behaviours are pulling in new directions. But you could call this phenomenon what it truly is: a mass secession from the truth, propelled by the failure of trust.

Pulsar starts the report by saying,

“We’ve been here before. In the middle of the 20th century, particularly in the US, markets and audiences were concentrated in the hands of a few major players across telecoms, automobiles and personal computing – until they weren’t. Today, the same fragmentation that took place within these industries is happening to social media.”

This ‘fragmentation’ could easily be seen as a visible symptom of ideological distrust. When institutions, governments, newspapers, and the very platforms that host them are viewed with a healthy scepticism, the result is predictable: a retreat. Platform choice is now a form of identity signalling. People are no longer merely choosing a medium; they are choosing a tribe, a belief, and a preferred version of reality.

Take Substack, for example. Pulsar highlights that Substack’s growth has been 65% this year – based on conversations here, this is in part due to a philosophical known that Substack rewards engagement and free speech.

The biggest user growth also comes from Bluesky and Truth Social – two other social platforms with strongly positioned purposes. In contrast, the user base for goliaths such as TikTok and Twitch are falling; Facebook and LinkedIn seem flat. It could be perceived that the masses, refusing to be informed by a single, purportedly objective source, now seek refuge in niche outlets and specialized channels where their existing prejudices are not challenged but celebrated.

Communications professionals with deep audience understanding will need to face how to manage narratives across these splintered grounds; additionally, where to focus shrinking budgets to make the most impact.

Get ready for 2026

If I had to summarise what 2026 will mean for strategic communicators, then it is to witness a paradox: we are more internet-connected than at any point in our lives, yet we are drowning in misinformation, polarisation, and isolation.

Social media has resulted in a society that is neither social nor particularly expert (thanks, AI). Having been burned by the cold, algorithmic indifference of the major platforms, where outrage is the currency, people now crave something that the corporate world, in its infinite clumsiness, struggles to provide: authenticity.

The fragmented communications landscape has curdled the professional tone of voice, which is likely suspected as PR-managed deception. Instead, people crave the personal. People do not want to hear from a “brand”; they wish to hear from a human being, flaws and all. It’s like the social media of 2008 has come full circle for 2026.

This means moving away from the vanity of “broadcasting content” and towards the essential work of community building. Trust is no longer a top-down endowment; it is a bottom-up achievement. It is built in the small –the Discords, the Substack comments, Facebook private groups – where the noise of the mob is replaced by the voice of the person.

The irony is delicious: to be successful in the professional world of 2026, one must become decisively less “professional.” One must be prepared to be personal, to be eccentric, and to be present. We are no longer managing reputations; we are tending to small fires of human fellowship in a very cold, very dark, and very splintered woods. If we cannot provide that warmth, we are merely adding to the chill.

Like many intentions of parents, they are quickly influenced by the actions of their children. In my case, another nappy change is now due. You could say I like to get my hands dirty, both professionally and in my role as dad. Clearly, part of my 2026 challenge will be balancing ‘corp dad’ with ‘dad, dad’.

Merry Christmas.

“… an abnormal and unstable situation that threatens an organization’s strategic objectives, reputation or viability”

Among client conversations and internal team discussions, I can tell you that the initial reaction to the financial violence isn’t fear, but a profound sense of offense. Without divulging the details of my first taste of short seller attack, often an organisation’s first reaction is the desire to drag the accuser into the public square and flay them with facts. For short sellers though, this is a mistake and a misjudgement of their character; they rely on narratives that drive a belief that a stock’s price is destined to fall.

Giving oxygen to this flame, say through an in-depth rebuttal, only acts to fuel the initial accusations. The objective is to lower the stock and then buy back shares at a lower price – a process known as “covering” – to profit from the difference. A short seller does not seek a debate; they seek a profit. Beyond financial value to reputational value, it only takes 10 seconds to smear a reputation, and 10 months to scrub the wall clean.

This is a grim theatre that began playing out last week when Grizzly Research, a short seller led by Siegfried Eggert, released a report accusing the FTSE 100 firm Trustpilot of operating a “mafia-style extortion campaign”. My own principles refrain me from directly linking to the report from this humble Substack, as arguments made are a classic of this short seller genre.

The assault on Trustpilot

The report was conspiratorial, laced with the sort of hyperbolic adjectives that usually signal a weak argument, collecting a small number of examples to claim Trustpilot actively courts fake negative reviews to bully businesses into paying for subscriptions; a “pay-to-play” racket. The examples provided weren’t representative of Trustpilot’s size: 330 million active reviewers and 60 million monthly active users.

Trustpilot responded within hours,

“The Report presents a series of claims that are selective, misleading and framed to support a predetermined narrative. It omits key context and publicly available facts, creating a false impression and exhibits a lack of understanding of how Trustpilot works. Trust is our guiding principle and is central to everything we do.”

More importantly, the directors put their own capital on the line. CEO Adrian Blair, Chair Zillah Byng-Thorne, and CFO Hanno Damm all bought shares in the immediate aftermath. This further signalled confidence in the business.

The Trustpilot response offers a textbook study in strategic communications during a crisis of confidence. The temptation, when accused of being a “mafia” operation, is to shout. As I wrote earlier, one wants to issue a 50-page line-by-line rebuttal. But the strategic communicator knows that to do so is to grant the aggressor the status of a peer. By responding to every point, you legitimise the claims. You accept the premise that you are on trial.

Trustpilot avoided this trap. They denied the central thesis with brevity and then went back to work. The share purchases were the real communication.

In my own experience, the most vital lesson is to distinguish between the noise of the market and the signal of the business. The short seller relies on the “short and distort” tactic: take a kernel of complexity (e.g. Trustpilot’s open platform allows anyone to post, which naturally includes some bad actors), strip it of context, and present it as a systemic fraud.

The antidote is not to argue the complexity, but to demonstrate the simplicity of the value. Trustpilot is the standard for online trust; if it were truly a criminal enterprise, one suspects its millions of users and the scrutiny of the London Stock Exchange might have noticed before Grizzly Research arrived.

Trust is reputation, especially when it’s in your name

The Trustpilot share price has already begun to rebound, suggesting that the market’s immune system is functioning. Trustpilot plays a significant role in portraying the perceived reputations of companies in Google Search and in Generative AI results – so this crisis strikes at the heart of their business model.

The long-term imperative for Trustpilot is to further shed light on how they operate, work to correct bad actors seeking to manipulate the platform, and demonstrate that businesses that pay receive support but not manipulation. Trustpilot is a business well-positioned for sustainable growth, even more so given how Generative AI is changing online journeys.

… and in case you’re wondering, Trustpilot is not a client. Well done to their communications team and any agency that may have been supporting.