The traditional search engine was built around a simple exchange: a user entered a question, received a ranked list of links and chose where to go next. That model is no longer the default. It is now competing with — and inside Google, being wrapped by — search products that answer questions directly, group results with AI and invite users to continue a conversation rather than begin a new search.
The clearest marker came on 19 May 2026, when Google used its I/O keynote to announce what it called a new era for AI Search. Gemini 3.5 Flash became the default model behind AI Mode for everyone globally. Google also introduced a redesigned search box, which it described as the biggest upgrade to that box in more than 25 years: it expands as you type, suggests how to phrase a question, and accepts images, files, videos and Chrome tabs as inputs alongside text. Users can now ask a follow-up question directly from an AI Overview and slide into a conversation in AI Mode without losing context. Google says its information agents will work in the background around the clock, monitoring the web for changes relevant to a standing question, and that Search will be able to assemble generative interfaces — tables, simulations, small trackers — on the fly rather than returning a page of documents.
The scale behind those features is Google's own reporting, and worth reading as such. In a June 2026 post aimed at website owners, the company said AI Overviews had passed 2.5 billion monthly active users and AI Mode more than one billion monthly users. At I/O it said AI Mode queries had more than doubled every quarter since launch, and that total Search queries had reached an all-time high.
The results page is now an interface
This does not mean the search engine is disappearing. Search has changed before: from a page of blue links to a mix of maps, images, videos, shopping results and answer panels. StatCounter still puts Google at close to 90% of global search referrals, a figure that has moved by roughly a point either way over the past three years. The UK's Competition and Markets Authority found the company handles more than 90% of search queries in Britain.
What is changing is the role of the results page. It is becoming an interface that interprets, summarises, recommends and increasingly acts, rather than one that points outward. Google's own documentation describes how AI Mode breaks a question into subtopics and searches each one simultaneously, then composes a response from what it finds — a different retrieval path from the ranked list underneath it.
Google is candid about the limits. Its help pages carry the line that AI responses may include mistakes, and state plainly that AI Overviews can and will make errors. The same pages note that AI Overviews are a core Search feature that cannot be switched off, though a Web filter still returns text links only.
What the evidence says about clicks
The concern that a concise answer makes the original source less visible has moved from argument to measurement.
Pew Research Center's clickstream study, published in July 2025 and based on the browsing data of 900 US adults, found users clicked a traditional result on 8% of visits to search pages carrying an AI summary, against 15% of visits to pages without one. Clicks on the links inside the summary itself occurred on 1% of visits.
That evidence was correlational. In April 2026, Saharsh Agarwal of the Indian School of Business and Ananya Sen of Carnegie Mellon published a working paper on SSRN reporting the first randomised field experiment on the question. Using a browser extension, they assigned 1,065 US desktop Chrome users to see search normally, to have AI Overviews hidden in real time, or to be routed into AI Mode. Where an AI Overview appeared, outbound organic clicks fell by 38% and the probability of a zero-click search rose by about a third. Sponsored clicks and overall search frequency were unaffected, which implies the loss is concentrated in organic traffic rather than in search activity overall. The authors also found no measurable improvement in how satisfied users were. The paper is a working draft and has not been peer-reviewed; a later revision reports a slightly larger effect.
Aggregate publisher data points the same way. Analytics from Chartbeat, cited in the Reuters Institute's 2026 trends report, showed organic Google search traffic to more than 2,500 sites down 33% globally between November 2024 and November 2025, and down 38% in the United States. Surveyed publishers expected search referrals to fall by a further 43% over three years. In July 2026, the Wall Street Journal reported traffic declines of more than 40% for some titles between June 2025 and June 2026, using Semrush figures — though not universally: the Guardian and the BBC gained over the same period.
From measurement to leverage
The more consequential change since this piece was first written is that publishers acquired something they did not previously have: a legal lever, and a credible threat.
On 3 June 2026 the CMA issued conduct requirements on Google under the Digital Markets, Competition and Consumers Act, following the company's designation with strategic market status in general search. Publishers must be able to opt out of having their content used in generative AI features without losing their position in ordinary search results, and to refuse the use of their content for model fine-tuning. Google has nine months to comply in full, with the regulator monitoring implementation.
Google moved the same day, announcing a Search Console toggle that lets site owners decide whether their content appears in and grounds responses in generative AI features, including AI Overviews, AI Mode and AI Overviews in Discover. Two details matter. Sites that opt out receive no traffic and no impressions from those surfaces. And Google states the setting is not used as a ranking signal outside them. Alongside it, the company began rolling out Search Console insights showing impressions in AI features and which pages appear in AI responses, in which countries. Both are being tested with a subset of UK site owners first.
The threat is the other half. According to the Journal, USA Today's parent has said it may cut off Google's access entirely if search traffic keeps falling; Reddit is reviewing its reported $60m-a-year arrangement; Politico and Reuters have weighed limiting crawler access and adding registration walls. Penske Media sued Google over AI summaries in September 2025. Cloudflare, which reports that bots now account for more than half of all web traffic, said it would begin blocking multi-purpose crawlers — those that scrape both for search indexing and for AI training — by default from 15 September unless site owners choose otherwise. What was once unthinkable, trading away search discoverability, is now a live board-level option.
Both stories are true at once
Set against this, Alphabet's numbers show no revenue cannibalisation. For the quarter ended 30 June 2026, Search and other revenue was $63.3bn, up 17% year over year, with Sundar Pichai attributing growth to AI features driving query volume and describing a single seamless experience across AI Overviews and AI Mode.
These findings are not in conflict. Total searches can rise while the probability that any one search produces an external click falls — and Google can monetise a larger pool of queries while sending a smaller share of them outward. That is the mechanism behind almost every disagreement in this debate.
The quality question
Speed is not the same as reliability. Research coordinated by the European Broadcasting Union and led by the BBC, published in October 2025, had journalists at 22 public-service media organisations in 18 countries assess more than 3,000 responses about the news from ChatGPT, Copilot, Gemini and Perplexity in 14 languages. It found that 45% contained at least one significant issue; 31% had serious sourcing problems such as missing or misleading attribution; and 20% had major accuracy problems, including invented details and outdated information. Gemini performed worst, with significant issues in 76% of its responses, driven largely by sourcing.
Audiences appear to sense this. The Reuters Institute's Digital News Report 2026, based on nearly 100,000 respondents across 48 markets, found weekly use of AI chatbots for news rose from 7% to 10% year on year, reaching 16% of under-35s — fast growth, but not yet a replacement. Trust in answers from AI chatbots stands at 20% globally, against 37% for news overall. Among those who do use chatbots for news, 42% say they always or often click through to the original source, sitting between social media (36%) and search (44%), though those figures rest on different user bases. The report also notes that Google mostly does not enable AI answers for hard news queries, a reminder that the traffic story differs sharply by category, and that utility content has been hit hardest.
For readers in markets like Nigeria, where the same report finds concern about false or misleading information among the highest anywhere and reliance on creators well above the global average, an interface that summarises confidently and sources loosely is not a neutral development.
What the question has become
The central question is still not whether people will stop searching. They are searching more. It is whether searching remains an act of exploration and, increasingly, who gets to decide.
If the answer arrives before the trail of sources, the power of search lies in how an interface frames knowledge, not only in what it helps users find. What has changed in 2026 is that this framing is no longer purely a product decision. It is being negotiated in regulators' conduct requirements, in crawler defaults set by infrastructure companies, in licensing terms, and in the calculation each publisher now has to make about whether visibility is worth the price of being summarised.
The route to evidence used to be a by-product of how search worked. It is becoming something that has to be argued for.
