Google AI Overviews Traffic Losses Deepen as Links Move Higher

Google tries to answer the traffic complaint

On October 15, Google said it will show links to websites and articles more prominently, and higher up, in its AI-generated search results. It plans to roll its link "carousel" out globally and to favor articles from sources that users have marked as preferred4. The timing reflects pressure on the company. The announcement landed alongside new evidence that AI Overviews, which now reach two billion monthly users, are steering clicks away from the sites they summarize4.

This is not Google's first fix. In May 2026 it added inline links, desktop hover previews, a "Subscribed" label and "Further Exploration" prompts to AI Overviews and AI Mode, all explicitly meant to send more traffic back to publishers5. Google's position is still that Search sends billions of clicks to the web every day, and that its AI features alone send billions every week9. Critics point out that this figure can't be checked. Google has given no baseline, denominator or timeframe, and Search Console reports only impressions for AI features, not clicks10.

The outlook is clear: moving links higher won't reverse what has already happened. The open question is whether it slows the decline.

How big is the publisher traffic loss?

The numbers differ by methodology, but they all point the same way. Chartbeat's 2026 Publisher Playbook found Google Search referrals to news and media sites fell 40.2% between July 2025 and July 20262. Across Chartbeat's network, Google Search dropped from 9% of all pageviews in July 2024 to 5% in July 2026, a 46% relative decline over two years2. An earlier Reuters Institute analysis of more than 2,500 sites found Google organic referrals down about 33% globally and 38% in the US between November 2024 and November 20253. Publishers in that study expect another 43% decline over the next three years5.

The damage is unevenly spread. Small publishers have lost about 60% of search referrals over two years, compared with about 22% for large ones1. Mid-sized publishers lost 47%10. Company results show the cost. UK publisher Reach said Google referrals fell about 55% and digital page views about 40%. Its half-year digital revenue fell from £61.1 million to £54.2 million, even though revenue per thousand page views rose 49%1. That combination makes the core problem plain: better monetization of each visit can't make up for a sharp drop in the number of visits1.

Advertising supply data shows the same thing. Ozone figures reported by Digiday showed premium publisher ad-request volumes in Q2 2026 down 32% to 37% year on year in the US and 39% to 41% in the UK8. Business Insider cut 21% of its staff after its search traffic fell. HuffPost lost more than half its organic search traffic over three years, and The Washington Post nearly half1.

Is AI to blame? The evidence is now causal

Until recently, Google could reasonably say that falling traffic had many causes. That defense is weaker now. A randomized field experiment by Saharsh Agarwal of the Indian School of Business and Ananya Sen of Carnegie Mellon followed 1,065 desktop Chrome users. When an AI Overview appeared, outbound organic clicks fell 39.8% and zero-click searches rose 34.5%, while clicks on sponsored results stayed flat610. Removing AI Overviews raised outbound clicks from 0.37 to 0.62 per search and cut the zero-click rate from 73% to 54%5.

Observational studies agree on direction, though not on size. Ahrefs measured the top result's click-through rate falling 34.5% when an Overview appeared. A February 2026 update widened that gap to 58%10. Pew Research found visits ended in a result click 8% of the time with an AI summary and 15% without one10. The more cautious numbers come from Amsive, which measured an average 15.49% drop in click-through rate and an 18.68% gain for branded queries10. Brand searches appear largely protected.

There are honest disagreements here. The Reuters Institute says it isn't clear how much of the aggregate decline comes from AI Overviews rather than algorithm changes or other factors5. Business Insider's three-year slide likewise can't be blamed on AI alone1. Google's Liz Reid has argued that total organic clicks are "relatively stable" and that reports of dramatic declines often rely on flawed methods10. Still, the experimental data shows that AI summaries cause a meaningful share of the loss, even if they don't explain all of it. Google's overall view now looks hard to square with that evidence.

AI chatbots have not replaced the lost traffic. Chartbeat puts AI referrals at about 0.01% of publisher traffic, and that share slipped during 20262. Similarweb reports that ChatGPT referrals are rising, but from a much smaller base8.

Courts close a door, regulators open another

Publishers' legal options in the US have narrowed. On September 30, 2026, Google won dismissal of both Chegg's antitrust suit and Penske Media's suit, without prejudice10. Penske, which owns Rolling Stone and Variety, had argued that its affiliate revenue fell by a third from its late-2024 peak10. As one analysis summarized the ruling, the judge found no agreement that obliged Google to send publishers traffic7.

Regulators are where the fight continues. On June 3, 2026, the UK Competition and Markets Authority ordered Google to let publishers keep their content out of AI Overviews without losing search rankings, and to cite publishers with clear links. Google has nine months to comply7. Google has resisted these opt-outs, warning they would harm Search5. In France, the newspaper group Apig told the Competition Authority in August that AI summaries violate 2022 agreements with news publishers4. Researchers describe publishers as stuck: opting out of AI summaries has meant opting out of Google search, which would hurt even more3.

Algorithm updates add more volatility

As AI Overviews reshape what users see, Google has kept changing its rankings underneath, and coverage of the 2026 update calendar doesn't fully agree. Several trackers count only two confirmed broad core updates, in March and May (the May update ran May 21 to June 2), plus spam updates in June and August2015. ALM Corp, by contrast, describes a late-June-to-July update that hit AI content farms and small affiliate sites hard11. Some outlets also describe a September core update that Google's dashboard doesn't appear to list1720. Given that disagreement, any traffic drop that a site blames on a specific "update" deserves skepticism.

On the overall direction, the reporting broadly agrees. Google says it targets scaled, low-value content rather than AI involvement itself12, and it has confirmed that its spam policies now cover attempts to manipulate or buy citations in AI Overviews and AI Mode14. The May update launched two days after Google I/O. There, Google said AI Mode had passed one billion monthly users with queries more than doubling every quarter, made Gemini 3.5 Flash the default model, and merged AI Overviews and AI Mode into one continuous experience1318. Late August brought another change: some AI Overviews now expand automatically into longer, AI Mode-style answers20. One tracker found only about 4.5% of AI Mode sessions end in a click to an outside site, compared with roughly 24% for traditional search20.

The link between ranking well and being cited is also unclear. Some practitioners say 99% of URLs cited in AI Overviews come from pages already in the organic top 1030. Others report the overlap fell from about 76% in mid-2025 to somewhere between 17% and 54% in early 202611. Only 14% of URLs cited in AI Mode reportedly also appear in AI Overview citations16.

Answer engine optimization grows up, and Google says it's still SEO

Marketers have responded with a new set of terms: answer engine optimization (AEO), generative engine optimization (GEO) and LLM optimization. There is no agreed definition21. Google's 2026 guidance mentions these terms but says optimizing for generative AI search is still SEO. Microsoft's Bing documentation uses "generative engine optimization" while stressing that SEO fundamentals still determine eligibility21.

The common playbook is consistent across guides. It calls for:

• direct answers placed early on the page
• headings phrased as questions
• lists and tables
• named authors and sourced statistics
• crawlable, fast-loading pages

The reasoning is that AI Mode breaks a query into many sub-questions and pulls in individual passages, not whole pages29. The academic GEO research these guides cite found that adding citations, statistics and quotations could raise visibility in generative answers by up to 40%25. Mentions on other sites also count heavily. One study found brands 6.5 times more likely to be cited through third-party sources than through their own domains23.

Not every claim in this growing industry holds up. First Page Sage, itself an AEO agency, says it has seen no measurable ranking effect from structured data, despite the many guides calling schema essential for AI visibility1828. One area does have hard evidence: brand matters. Branded queries keep their clicks10, and Google is now rewarding sources that users choose as preferred4.

The takeaway

The evidence points one way. AI answers are permanently shrinking the open web's share of search clicks, and Google's link tweaks only adjust around the edges. For publishers, the lesson from Reach and Business Insider is to build direct relationships with readers, through subscriptions, loyal audiences and recognizable brands, rather than chase citations12. For marketers, AEO is worth doing, but it amounts to SEO plus brand-building. Neither group should count on chatbot referrals or US courts to restore the old traffic.