The click is being repriced: what AI search is doing to the open web

In early November 2026, AdExchanger reported that People Inc.’s referral traffic from Google search had fallen 70 per cent over two years and 22 per cent in the third quarter, after a 40 per cent fall in the second. The report was corrected the next day to separate search referrals from core sessions; core sessions fell 26 per cent. Chief executive Neil Vogel said AI Overview summaries appeared at the top of 90 per cent of Google searches associated with its content. The traffic loss contributed to 195 layoffs, about 6 per cent of staff.

On the same call, digital revenue rose 2 per cent year on year to $289 million, the twelfth consecutive quarter of growth, although total revenue for the media business fell 5 per cent to $409.9 million. People Inc. was losing the search traffic that had long fed its business, yet it was still finding digital revenue elsewhere.

That contrast frames the question better than claims that the internet is dying. The evidence does not describe a clean collapse. It describes a re-pricing of attention: search clicks are becoming less dependable, replacement channels are still small, and the financial damage depends on what a publisher can sell besides a visit.

The bargain under strain

Search has worked as a referral system. A reader searched, a result earned a click, and a publisher had a chance to earn advertising or subscription revenue and build a direct relationship. AI summaries interrupt that chain by answering some queries on the results page, so readers can be satisfied without making the trip.

The strongest evidence for click diversion is a randomised field experiment described by Digital Content Next and PPC Land. Saharsh Agarwal of the Indian School of Business and Ananya Sen of Carnegie Mellon University used a custom browser extension to vary whether 1,065 US desktop Chrome users saw AI Overviews. The analysis covered 68,089 unique searches, with overviews triggered in roughly 41 per cent of queries. When an overview appeared, outbound organic clicks fell 39.8 per cent and zero-click searches rose 34.5 per cent.

The design gives the result its force. For queries where an overview was meant to appear, hiding it raised outbound organic clicks from 0.37 to 0.62 per search and cut the probability of a zero-click search from 0.73 to 0.54. Sponsored clicks and clicks within Google’s own properties did not move, and queries without an overview showed no difference between groups. That points to AI Overviews displacing organic visits, not merely rearranging attention across the page.

The experiment also tested Google’s quality-click defence, which holds that fewer clicks may be better clicks. It found no meaningful difference between groups in back-button navigation, a bounce measure defined as a session under ten seconds with no further interaction, or total time on page. An endline survey found no difference in satisfaction, perceived information quality or ease of finding information. Those are averages, not proof that every quality claim is false, but the defence cannot carry the argument alone.

The experiment’s details come second-hand through Digital Content Next and PPC Land. Its sample is US desktop Chrome users, and it measures clicks, not revenue, subscriptions or long-term publisher health. It shows that an AI Overview can divert an organic click, not what that means financially.

Digital Content Next reported that Pew Research Center observed a similar pattern in an analysis of nearly 69,000 searches by 900 US adults: users clicked a traditional result on 8 per cent of visits when an AI summary appeared, compared with 15 per cent when one did not. Links inside summaries drew clicks on 1 per cent of visits.

Traffic loss, revenue mix

People Inc.’s results show the gap between clicks and revenue. Non-session revenue, not tied to visits, made up 43 per cent of digital revenue and grew 15 per cent. Session-based revenue, the remaining 57 per cent, fell 6 per cent. Licensing revenue grew 49 per cent, led by deals with aggregator platforms and AI vendors. Advertising revenue still fell 8 per cent, even as higher premium ad rates partly offset lower traffic.

This is adaptation, not immunity. Digital growth was slower than the fall in search referrals, and advertising, the line most exposed to visits, kept shrinking. The company expected licensing growth to slow in the fourth quarter. Non-session lines may be less predictable than a broad stream of search visitors, and one company cannot represent an industry. But it rules out the simplest collapse story: a publisher can lose most search referrals and still report digital revenue growth.

Google’s counter-case

Google’s position is the main counterweight. The Indian Express reported in August 2025 that Google called third-party findings on AI search and web traffic inaccurate and dramatic, saying they rested on flawed methodologies, isolated examples or traffic changes that predated AI features in Search. Liz Reid, its vice president and head of Search, said organic click volume from Google Search to websites had been relatively stable year on year and average click quality had increased. The report noted that Google did not publish figures to support those claims.

The Indian Express also reported Google defines a quality click as a visit where the user is interested in the site, indicated by not immediately clicking back. Reid argued that users will still click through when they want to explore a topic more deeply or make a purchase. Some queries end with the answer; others lead to research or buying, so the claim is plausible in principle. But Google’s account is self-reported, predates the 2026 randomised study and was not backed by published figures in the selected report.

The Indian Express reported that Similarweb found AI platform referrals to news and media sites up 770 per cent from June 2024, with 1.13 billion to the top 1,000 sites in June 2025 and ChatGPT taking 80 per cent. Rapid growth from a small base can coexist with AI remaining a minor source.

The replacement channel is still small

Digital Content Next reported that the Reuters Institute, using Chartbeat data from 2,576 publisher sites, found Google Search referrals fell 33 per cent globally between November 2024 and November 2025, including 38 per cent in the US. Its researchers cautioned that AI Overviews could not account for all of the decline because algorithm changes, publisher strategies and audience behaviour also played a role.

In the same dataset, Google Search accounted for 7.3 per cent of publisher referrals and Google Discover 13 per cent. ChatGPT accounted for 0.02 per cent and Perplexity 0.002 per cent. Digital Content Next also reported that Sensor Tower found AI assistant traffic grew 86 per cent in 2025 while generative AI referrals were about 0.7 per cent of overall web traffic. Growth rates and market share must be read together.

IAB Europe added a cautious European view in March 2026: publishers across Europe reported declines in traffic and advertising revenue after AI chatbots and search summaries arrived. But it said there was no quantitative consensus on scale or economic impact, that evidence ranged from self-reported publisher metrics to aggregated analytics, and that traffic changes might reflect evolving behaviour as well as displacement. It called its work early and directional. AI referrals remained marginal, and losses appeared to exceed AI-generated traffic.

These sources do not agree on the size of the effect. The Reuters Institute measures a decline in search referrals, not AI causation. Chartbeat’s shares come from a particular publisher panel. IAB Europe covers Europe and calls its data directional. Similarweb’s growth figure and the small referral shares come from different periods and cannot be combined.

The monetisation gap

Forrester’s analysis on 29 October 2026 adds a platform-side twist. It said investment in conversational search was fuelling Google’s fastest year-on-year search query growth in five years, while Google’s advertising revenue growth slowed to 12 per cent, its slowest rate in five quarters. Search revenue was still growing. The issue was that query growth and monetisation growth were beginning to separate, even as referral traffic to the wider web came under pressure.

Forrester also reported Reddit chief executive Steve Huffman saying Google search referrals declined meaningfully in the third quarter because users stayed within AI Mode and AI Overviews. Reports from the earnings call put the October decline at more than 1 million referrals per week on average. That extends the concern beyond one publisher, though it remains a company account.

Google’s response is more advertising in the new surfaces. Forrester reported plans to scale ads in AI Mode and AI Overviews, expand them internationally and launch multi-turn ads in the US. Google says click-through rates are similar between sponsored and organic results in AI Mode. Forrester warns that heavier ad loads could erode trust, citing its own data that 70 per cent of US and UK answer-engine users trust the answer. That survey’s methodology is not visible, and the warning describes a risk, not a measured outcome. The tension is real: answers depend on trust, and ads can blur the line between answer and promotion.

What remains unresolved

The evidence supports a narrower conclusion. AI Overviews can reduce outbound organic clicks. AI referrals are not yet a replacement channel at the scale suggested by their growth rates. The financial outcome is mixed and depends on revenue mix: licensing, events and other non-session lines offer a partial offset while advertising remains exposed.

It does not establish that AI is the sole cause of publisher traffic declines or an industry-wide revenue collapse. It does not measure whether extra clicks produce subscriptions, purchases or loyal readers. Nor can it say whether licensing deals with AI companies will remain a durable substitute, especially when the same companies are building the answer surfaces that reduce referrals.

The practical question is not whether the internet ends. It is who pays for the material that makes an answer useful. If readers stop at a summary, publishers need revenue that does not depend on the visit. Licensing, events and other non-session revenue can provide some of it. None has yet proved it can fund the reporting and advice that answers draw on at the scale the old click economy did. The referral bargain is being rewritten. The evidence shows the strain, but not the ending.

A referral fall of 70 per cent and a referral share of 0.02 per cent can both be true of the same shift, which is why the base matters as much as the headline figure. Cybersecurity fundamentals builds the habit of asking what a number measured, over what period and for whom, so a growth rate is not mistaken for a market share and a self-reported claim stays labelled as one. That discipline travels well, and it is the same reasoning that keeps an observed effect apart from its presumed cause when two credible accounts disagree. To work through it yourself, explore The Hack Academy’s online cybersecurity training programme, where cybersecurity fundamentals is a listed subject area, with self-paced learning and practical exercises alongside the theory.