By Ronny Golan CEO and Co-Founder at ViewersLogic
New research suggests ChatGPT’s heaviest users are quietly abandoning Google, browsers, and news sites — raising urgent questions for publishers and advertisers alike.
There’s an irony buried in the AI revolution that nobody wants to talk about.
Large language models were trained on the open web (much of it funded by advertising), with billions of pages of journalism, reference content, and human knowledge. Now, new behavioural data suggests those same models may be quietly dismantling the ecosystem that made them possible.
That’s the uncomfortable conclusion lurking inside new research from ViewersLogic, a media measurement company that tracks real-world consumer behaviour through passively collected mobile data (what people actually do with their phones).
The findings are striking.
What the data shows
More than one in six UK mobile users (17.6%) used ChatGPT in September 2025. Most were casual dabblers; under 30 minutes for the entire month. But a distinct cohort of heavy users is emerging, and it’s within this group that the behavioural shifts become alarming.
On search: Once users crossed roughly one hour of ChatGPT usage per week, their weekly Google search volume dropped by 19%.
On browsers: Among heavy ChatGPT users, average weekly browser time fell from three hours and 25 minutes in January 2025 to two hours and 18 minutes by October — a 32.7% decline in under a year. The browser is quietly being replaced.
On news: Heavy users showed a 39% drop in time spent on news and newspaper websites in the first half of the year alone. For publishers already running on razor-thin margins and programmatic CPMs that barely cover the electricity bill, this is not a small number.
The self-defeating loop
Here’s where it gets philosophically thorny.
LLMs are trained on the open web. They depend on a constant supply of fresh, high-quality content — journalism, analysis, reference material — to stay relevant and accurate. That content is overwhelmingly produced by ad-funded publishers. But if AI tools are diverting users away from those very publishers, the economic model that sustains content creation starts to collapse.
Less traffic means less ad revenue. Less ad revenue means fewer journalists, fewer articles, fewer sources. Fewer sources means worse training data. Worse training data means worse AI.
We’re seeing the first real signs of AI becoming a front door to the internet for a subset of users. If AI steadily diverts usage away from the very news and reference sites it relies on for training data, the whole model risks becoming self-defeating.
Who’s actually using this stuff (and how)
The usage data is equally revealing. Among people who used AI apps in September 2025, 87% used only one platform. ChatGPT dominates that single-platform behaviour, though Perplexity is gaining ground as a primary research tool. Gemini usage also surged after the research period, ViewersLogic noted, suggesting Google’s tool has gained particular traction with developers.
And here’s something that cuts against the narrative that AI is purely a Gen Z phenomenon: while 18-24 year olds show higher penetration rates, it’s older professionals who are the heaviest users once they adopt the tools. Adoption is broad among younger cohorts, but intensity is higher among older ones.
That matters for advertisers. The heaviest AI users (the ones most aggressively abandoning traditional search and news consumption) may disproportionately be the high-income, high-intent professionals that brands have always paid a premium to reach.
What this means for the industry
The implications cut three ways.
For publishers, the trend raises the challenge: are AI platforms sending you traffic, or quietly absorbing the attention your content once captured?
For advertisers, the key question becomes ‘what happens inside the LLM?’ When a user asks ChatGPT for a product recommendation or a buying decision, what shapes the answer? There are no ad slots, no search rankings, or click-throughs to measure. Which is why the strategic priority is shifting from SEO to GEO (Generative Engine Optimisation) — brands that haven’t started thinking about how they show up within AI-generated responses are already behind.
For the AI platforms themselves, the sustainability question is real. The open web is not an infinite resource. It requires economics to function.
Our research is early-stage — these are signals, not verdicts. But the implications for advertisers, publishers and platforms will be significant.
The web taught AI everything it knows. Whether AI returns the favour is the defining media business question of the next decade.
ViewersLogic’s research is based on passively collected single-source mobile behavioural data from thousands of UK consumers.
