Google AI Search in 2026: AI Mode Push and Publisher Traffic Losses

A year of Google turning search results into answers

In 2026, Google search stopped being mainly a list of links and became an answer engine. The changes did not come as one launch. They built up through model swaps, interface redesigns, ranking updates and rulings from regulators. For publishers, each change made it less likely that a search would end with a visit to their site.

The year's most important change happened quietly. On January 27, Google reportedly put Gemini 3 behind AI Overviews and AI Mode without announcing it to site owners, and one analysis found that the number of sources cited in each AI answer rose by about a third.28 Two broad core updates followed. The March update ran from March 27 to April 8 alongside a spam update. The May update began on May 21, the second core update and fourth confirmed ranking change of the year.2322 The largest product change came at I/O on May 19. Google made Gemini 3.5 Flash the default model in AI Mode worldwide and introduced a rebuilt, AI-powered search box that it called its biggest upgrade in more than 25 years.21

The AI Mode default question: coverage diverges

The year's coverage disagrees most on one question: did AI Mode become Google's default search experience? Several marketing timelines say it did at I/O.2824 Google's own post says something narrower. Gemini 3.5 Flash became the default model inside AI Mode, and the redesigned search box encourages longer, conversational queries.21 Press Gazette's reporting a week later backs up the narrower reading. SEO consultants it spoke to described a "small sigh of relief" that AI Mode was not yet the default, while warning that the new features would still cut organic traffic across the board.4

The more accurate account is that Google made switching between modes almost effortless. Industry roundups describe a single flow in which a user moves from a standard results page with an AI Overview straight into a full AI Mode conversation, with no separate step.23 One SEO manager called this a trap: the classic results page handles the first query, and any follow-up question pulls the user into a loop where nobody clicks out.4 Formally, AI Mode is optional. In practice, Google seems to be making it the place where searches end up anyway. The company's own figures point the same way. It said follow-up queries in AI Mode rose more than 40% month over month, which suggests searches are turning into multi-step conversations.27

The reach is very large. Google reported more than 2.5 billion monthly users for AI Overviews and more than 1 billion for AI Mode at I/O 2026.9 Time noted that the AI Mode figure is comparable to ChatGPT's estimated audience, and cited Similarweb data showing AI Overviews on 40% of U.S. Google searches.10 Other trackers give different numbers. Comscore put AI Overviews on 39.4% of U.S. desktop searches by June, up from 25.8% a year earlier.28 Another estimate, citing BrightEdge and Ahrefs, puts the share near 48% of all queries.24 The exact share depends on the panel and the type of query, but every tracker shows it rising.

What the click data says

Research on publisher traffic has now moved beyond correlation. The most widely quoted figure comes from Ahrefs: a 58% drop in click-through rate for top-ranking pages when an AI Overview appears, almost double the 34.5% it measured in April 2025.2 Pew Research found that users clicked a traditional result 8% of the time when an AI Overview was shown, compared with 15% when none was shown. Users clicked a link inside the Overview itself in only 1% of cases.7

The strongest evidence came from a randomized field experiment first posted in April and revised in June. It followed 1,065 desktop Chrome users and found that showing an AI Overview cut outbound organic clicks by 39.8% and increased zero-click searches by 34.5%. Clicks on sponsored results stayed flat.8 Two further findings matter. The effect was concentrated almost entirely in cases where the Overview sat at the top of the page. And bounce rates and time on site did not differ between groups, which weakens Google's argument that the clicks AI does send are of higher quality.8 Google has said those clicks keep users on sites longer, but it has not published metrics publishers can check.1

The overall traffic numbers point the same way. Chartbeat data reported by Press Gazette show roughly 33% fewer search referrals to publishers worldwide since AI Overviews rolled out widely, and 38% fewer in the U.S.3 Size matters a great deal. Over two years, small publishers reportedly lost about 60% of their search referrals, mid-sized ones 47% and large ones 22%.5 Similarweb data analyzed by Axios show U.S. visits to about 100 top news sites falling from more than 70 million in 2024 to 47.6 million in July. Time noted that AI cannot account for all of that drop.10

There are some signs the decline is slowing. One dataset showed organic CTR recovering from a low of 1.3% in December 2025 to 2.4% in February 2026.3 Branded queries are the clearest exception: they show an 18% increase in CTR when an AI Overview appears.3 These exceptions do not change the overall result. They do show who benefits. AI search moves attention toward brands users already recognize, and the Search Engine Journal analysis warns this risks creating a two-tier web.1

Google's concessions and the regulatory turn

Google responded on May 6 with five changes to AI Overviews and AI Mode. They included a "Further Exploration" section of links, labels for news outlets a user subscribes to, more context around inline links, and previews of forum posts.2 The changes came as Penske Media pursued an antitrust lawsuit and the European Publishers Council filed a complaint with the European Commission.2 In February, Penske's court filing described Google as "cannibalizing" publisher traffic.7

Regulators did more than the voluntary fixes. On June 3, the UK Competition and Markets Authority required Google to let publishers opt out of having their content used in AI Overviews and AI Mode without losing their normal search visibility. It called the measure a world first. Main controls are due by December 2026, and page-level controls by March 2027.3 One timeline notes that a Search Console control for AI features arrived roughly ten weeks after Google told the regulator it was considering one, and only after the CMA made it binding.28 The lesson is that Google offers publishers concessions when regulators require them.

Core updates in an AI-first results page

The ranking updates matter in a different way now. Coverage of the May update describes it as rewarding first-hand, experience-based content and penalizing generic listicles. That matches Google's guide to optimizing for AI search, published on May 15.27 Reports differ on when the rollout finished: some give June 2, others June 4.2730 One tracker dates the update from May 4 and also lists an August core update tied to AI Overview citations.26 No other coverage reviewed here confirms either claim, so both should be treated with caution.

One practical point appears across the coverage: rankings and traffic have separated. SISTRIX data show the click-through rate for the first organic result on queries with AI features falling from around 27% to as low as 11%. A site can keep its ranking through a core update and still lose traffic.30 Measuring the effect is also harder because Google combines AI feature appearances with standard Search Console reporting.6

Answer engine optimization grows up, with mixed signals

That gap explains the rise of answer engine optimization (AEO) and generative engine optimization (GEO). Practitioners increasingly treat AEO and GEO as one discipline under two names.1219 Some guides still separate them, describing AEO as making content easy to extract and GEO as getting AI systems to cite it.1311 The research usually cited is the Princeton-led GEO study. It found that adding quotations, statistics and citations increased a source's share of AI-generated answers by roughly 28% to 41%, and that keyword stuffing had no meaningful effect.11

The main argument for optimizing for citations is that being cited may count for more than ranking. Pages cited inside an AI Overview reportedly get about 35% more clicks than a standard first-place result, though that figure comes from aggregated industry data rather than a controlled study.24 The overlap between AI Overview citations and the organic top 10 has also reportedly fallen from 76% in mid-2024 to between 17% and 38% by February 2026. If that holds, ranking well no longer reliably earns a citation.17

Google's own messages are inconsistent. Its AI search guidance says llms.txt files are not needed, but Lighthouse 13.3 now checks for one by default.22 Separately, one tracker says Google has started penalizing manipulative GEO "prompt-injection" content.26

The reading

The 2026 timeline shows a clear direction. Google is steadily shifting search toward answers produced on its own pages, and it is giving publishers some control mainly where regulators require it. The causal evidence now supports what publishers saw in their analytics. AI Overviews reduce clicks, and the remaining clicks are no better than the ones lost. AEO gives brands a way to stay visible inside these answers, but it does not bring back the traffic lost to them. The publishers most likely to cope are those that build direct relationships with readers through newsletters, apps and subscriptions, rather than relying on Google's AI answers to send visitors.1