3.24B Now Use Generative AI — the Open Web Is Bleeding

By Daniel Rao | 21 October 2026

One in three people on the planet now uses generative AI every single month. Three years ago that sentence would have read like science fiction; this morning it arrived, entirely matter-of-fact, in a San Francisco press release. And tucked inside the same report is something close to a casualty list: the open web — the thing this publication lives on — is losing traffic, fast.

Meltwater and We Are Social published their annual Digital 2027 report today, and it is one of those documents that makes you put your coffee down. The headline number — 3.24 billion active users of generative AI tools — is the kind of figure that forces a double-take. That is 38.9 per cent of humanity. Not of internet users. Of humans.

We checked the Digital 2027 press release against its Globe Newswire relay and two launch-day write-ups, and every headline figure matched the release line for line. The numbers are estimates — more on the caveats later — but they are consistent, and they tell a story nobody in publishing can afford to ignore.

Photo: A person uses a mobile phone to photograph a railway bridge near Essleben, Germany — mobile-first behaviour is exactly what the Digital 2027 data describes. Credit: Olaf Simons, CC BY-SA 3.0, via Wikimedia Commons.

Update, Oct. 21: Within hours of the press release landing, the trade press had picked its frame. MadTechMag's Oct-21 coverage headlined the report as "Over 1 in 3 People Now Use AI as Brands Face a Fierce New Battle for Attention" — reading Digital 2027 not as an AI story but as a marketing story, with the 3.24-billion-user figure, the 57.2% twelve-month growth, and the 5.81-billion social-media-identities milestone presented as the numbers advertisers must plan around. That framing rhymes with the release's own emphasis: the noticeable decline in traditional web traffic is attributed to users spending more time on social platforms, pivoting to native mobile apps, and embracing AI-powered platforms (Meltwater press release, Oct. 21). The report lands at an awkward moment for the web economy — our post's open-web analysis stands, and the first-day coverage suggests the brands will be asking the same survival questions as the publishers.

One in three humans now asks a machine instead of a person

The report estimates that active users of generative AI tools have grown to 3.24 billion, up an eye-watering 57.2 per cent in twelve months. That is 1.2 billion new users in a single year — roughly the population of India deciding, more or less at once, to start chatting with machines. For context, the report puts total global internet users at 6.15 billion, or 73.9 per cent of the world's population. More than half of everyone online is now, by this measure, an AI user.

The smartphone needed the better part of a decade to reach this kind of penetration; generative AI has gone from research curiosity to a habit of one-third of humanity in about three years. Whatever you think of the technology — a debate we have covered in the AI bubble debate — the adoption curve itself is now beyond argument.

But "active user" is doing a lot of heavy lifting in that sentence, and we will come back to that. First, the horse race.

ChatGPT is still the king — but Gemini is sprinting

ChatGPT remains the market leader by a distance: roughly twice the monthly active users of its nearest rival. OpenAI's first-mover advantage was not a fluke of timing. It was a moat, and it is still holding.

The chasing pack, though, is moving at a genuinely alarming pace. Active users of Google's Gemini mobile apps have increased fourfold over the past year, and Gemini's web audience is now more than 60 per cent the size of ChatGPT's. Anthropic's Claude is gaining ground too, though the report is lighter on specifics there. Read between the lines and the message is clear: the market is no longer a one-horse race, and Google's distribution advantage — a Gemini app pre-installed on billions of Android phones — is starting to bite.

This matters for a reason that goes beyond corporate bragging rights. Where people ask their questions determines who gets to answer them — and who gets the advertising revenue, the data, and the cultural influence that follows. A world in which AI assistants mediate a third of humanity's information diet is a world in which the companies that own those assistants wield extraordinary power. Which is precisely why the report's next finding should worry everyone who publishes on the open web.

The open web is bleeding out

Here is the number that should ruin a few afternoons in media boardrooms: across the world's 10,000 largest web domains, total monthly traffic has fallen by more than 10 per cent since the public launch of ChatGPT. The total time people spend using the web has dropped 11.2 per cent in just the past 24 months. The report frames it as a "significant shift in online behaviour" — users spending more time on social platforms, pivoting to native mobile apps, and embracing AI-powered platforms that answer questions without ever sending anyone to a website.

To be fair, the authors do not pin the entire decline on AI — "AI isn't the only reason web behaviours are evolving," the release notes. Apps have been eating the web for a decade. But the timing is brutal: the steepest part of the decline lines up almost exactly with the arrival of assistants that summarise the web so you don't have to visit it.

Publishers are already feeling it in their bones. Reach blaming Google's AI search for falling traffic was one of the first public admissions, but newsroom whispers suggest it will not be the last. When an AI assistant answers your question directly, the website that researched, wrote, and fact-checked the answer gets nothing — no visit, no ad impression, no subscription nudge. Search engines at least sent you the traffic; the new intermediaries keep the audience and give publishers a citation if they're feeling generous.

There is a grim irony here: the AI models were trained on the open web they are now hollowing out. Kill the incentive to publish, and the training data of the future dries up.

Social media swallowed another 151 million people

While the web shrinks, the walled gardens keep growing. Global social media user identities now stand at 5.81 billion — 69.9 per cent of the world's population — after adding 151 million in a year, a rise of 2.7 per cent. (Note the careful wording: "user identities", not people — one person with three accounts counts three times.)

More striking than the headcount is the time. The typical internet user now spends 19 hours and 18 minutes a week on social media and online video platforms, up 3.8 per cent on last year. That is nearly three hours a day, every day, scrolling, watching, and messaging. YouTube captures the largest share of total social media time worldwide, followed by WhatsApp and Instagram. But the intensity crown goes to TikTok: the average active user spends 1 hour and 55 minutes a day on it. One app. Nearly two hours. Daily.

Photo: A smartphone displaying a social media feed — Digital 2027 says the typical internet user now spends 19 hours and 18 minutes a week on social and video platforms. Credit: C.Suthorn, CC BY-SA 4.0, via Wikimedia Commons.

The generational split is where it gets interesting for anyone trying to sell something. Among internet users aged 16 to 24, 46.9 per cent now use platforms like Instagram and Facebook as their primary source of information when researching brands and purchases. Not Google. Not review sites. Social feeds. Mobbie Nazir, Global Chief Strategy and Growth Officer at We Are Social, put it bluntly in the release: "The data in Digital 2027 proves that we are experiencing a fundamental shift in how people behave online. With social media use reaching 5.8 billion and gen AI woven into the daily habits of over a third of the world, the competition for attention is fiercer than ever."

His prescription — brands must "go out and actively meet consumers where they are" — sounds like marketing speak until you realise it is also a survival guide for publishers. The audience has moved; the question is whether the business model can follow.

Search engines are still beating AI at one thing: selling stuff

Before the AI triumphalists get carried away: despite AI's momentum, online adults are still 74 per cent more likely to say they visit search engines when researching brands than to use an AI platform. For all the talk of ChatGPT replacing Google, when money is about to change hands, people still want the blue links — or say they do.

The typical online adult now relies on an average of 5.8 different channels to discover new brands, and the money is following the fragmentation. Digital ad spend rose 8.7 per cent over the past year, with social media advertising up 18 per cent and influencer ad spend up 15 per cent. Alexandra Bjertnæs, Chief Strategy Officer at Meltwater, captured the new reality in the release: "Audiences no longer discover, research or make decisions about brands in one place. They move across news, social platforms, mobile apps and, increasingly, AI assistants. With generative AI adoption growing nearly 60% in the last year alone, LLMs have quickly become an environment where first impressions are formed."

That phrase — "an environment where first impressions are formed" — is doing quiet, devastating work. It means the AI assistant is becoming the front door. And whoever controls the front door controls what billions of people believe is true about your company, your product, or your politics. For a deeper look at how these autonomous systems actually work under the hood, see our AI agents explained piece.

Why you should read these numbers with one eyebrow raised

Now the caveats, because a report this dramatic deserves scrutiny. Meltwater sells social-listening tools; We Are Social sells social media marketing. Neither has a commercial interest in telling you the open web is thriving. That does not make the numbers wrong — the report draws on GWI, Similarweb, and Statista across 900 pages — but the framing serves a purpose. Always ask who benefits when a statistic is this convenient.

Then there is the definition problem. "Active user of generative AI tools" is the report's own construct, and the release does not publish the survey methodology in the summary. Does someone who asked ChatGPT one question in March count the same as a daily power user? Almost certainly. The 3.24 billion figure is best read as a ceiling — the number of humans who have meaningfully touched the technology — not as 3.24 billion devoted converts.

The release also offers no US-specific split. Adoption curves run hotter in some regions than others — ecommercenews.asia notes that 48.7 per cent of Singaporeans aged 16 and above used ChatGPT in the past month, well above the 34.7 per cent global average. The global figure smooths over enormous regional variation, and says nothing about frequency or whether people trust what the machines tell them.

And yet — even heavily discounted, the direction of travel is unmistakable. You do not get a 57 per cent annual growth rate, a 10 per cent web-traffic decline, and 5.81 billion social identities from statistical noise. The exact numbers are arguable. The trend is not.

Oh, and the report's lightest finding, buried among the charts like a reward for reading to the end: Google search data confirms cats continue to rule the internet at the end of 2026, fetching 4.15 billion results compared to 4.02 billion for dogs. Some things, mercifully, never change.

The colossally uncomfortable truth for publishers

Here is where I stop summarising and start editorialising, because this story lands uncomfortably close to home. This publication — like every publication — lives on the open web the report says is bleeding out. More than 10 per cent of traffic gone from the web's biggest domains since ChatGPT launched. Web time down 11.2 per cent in two years. And the replacement for that traffic is a set of AI assistants that answer questions using our work while sending us nothing in return.

The industry's response has been colossally inadequate: a few lawsuits, some licensing deals that amount to hush money, and hopeful talk about "AI visibility optimisation" — as if the answer to being strip-mined is to make yourself easier to mine. Google paying publishers for AI search content was at least an admission that the arrangement is shoddy. Most AI companies have not managed that much honesty.

But the doomsayers miss something: the audience has not stopped being curious — it has moved rooms. Nineteen hours a week on social and video; nearly two hours a day on TikTok. People are not reading less; they are reading differently, in feeds and chats and AI answers rather than on homepages. The publishers who survive will be the ones who learn to be indispensable inside the feed, the chat, and the answer itself. That is a much harder job than publishing. It is also the only one on offer.

Frequently asked questions

Is ChatGPT still the biggest AI tool?

Yes, comfortably — roughly twice the monthly active users of its nearest rival. But Google's Gemini is closing fast: its mobile app users quadrupled in a year and its web audience is now over 60 per cent of ChatGPT's size. The lead is real; the moat is narrowing.

Does rising AI use mean people will stop visiting websites?

Not entirely, but the trend is ugly for publishers: traffic across the world's 10,000 largest web domains is down more than 10 per cent since ChatGPT's public launch, and time spent on the web has fallen 11.2 per cent in two years. AI assistants, social feeds, and native apps are absorbing attention that used to go to the open web.

Is the Digital 2027 report reliable?

It is serious — 900 pages drawing on GWI, Similarweb, and Statista — but also a vendor report from companies selling social-listening and social-marketing services, so read the framing sceptically. "Active user" is the report's own definition, and there is no US-specific breakdown. Treat the figures as well-informed estimates, not census data.

Are these figures for the US or the whole world?

Global. The release gives worldwide totals with no country-level split. Regional behaviour varies enormously — Singapore reports 48.7 per cent ChatGPT usage among over-16s against a 34.7 per cent global average — so US-specific numbers would almost certainly look different from the global mean.

What does Digital 2027 mean for brands and advertisers?

The press release frames it as an attention shift, not a tech story — people now research brands and purchases on social feeds. As the post notes: "Not Google. Not review sites. Social feeds." Mobbie Nazir's prescription is that brands must "go out and actively meet consumers where they are". Trade press picked up the same signal — MadTechMag's Oct-21 coverage headlined the report as brands facing "a fierce new battle for attention". The audience has moved; the business model question is whether brands and publishers move with it.

Daniel Rao covers AI, platforms, and the business of the internet for TechPulse USA, where he writes about the collision between new technology and the economics of attention.

Sources: Meltwater / We Are Social "Digital 2027" press release (San Francisco, 21 October 2026); Globe Newswire relay via Business Intelligence; ecommercenews.asia; TechSabado. Figures are the report's estimates; see caveats above.