Google shifts search to AI mode

Google used its May 19 I/O conference to push Search further toward chat-style answers, adding AI Mode as a more conversational interface and tying more search experiences to Gemini. The change extends a shift already underway with AI Overviews, but AI Mode goes further by inviting follow-up questions and longer back-and-forth exchanges inside Search. (blog.google) Google says the goal is to help people explore the web more deeply; critics say the format could keep users inside Google’s interface longer and send less traffic to publishers. (support.google.com) Google’s own support page carries a warning that AI Mode can make mistakes. That caveat matters because the company is now asking users to rely on a system that summarizes, selects and sequences information before many people click through to source pages. (support.google.com) At the same time, reporting by the BBC said Google and other AI systems have faced attempts to manipulate generated answers, prompting Google to update its spam policies. ### What exactly changed in Search? Google said at I/O 2026 that AI Mode is part of a broader overhaul of how people search, with Gemini integrated more deeply into results and follow-up questions. The company’s blog says AI Mode and AI Overviews are being updated to make it easier for users to “dive deeper” and continue exploring after an initial response. (blog.google) Google’s help documentation describes AI Mode as an experimental Search feature available through Search Labs for early access users. The interface is designed for multi-step queries and conversational back-and-forth, rather than the older pattern of typing a query, scanning links and reformulating from scratch. (support.google.com) ### Why are publishers and web sites worried? Business Insider argued the new setup could reduce referrals by answering more questions directly inside Google, and TechCrunch described the broader redesign as a version of Search that no longer looks primarily like a list of links. Google, for its part, says it is adding more ways for people to reach websites, including suggested next-step links and source connections inside AI responses. (blog.google) The dispute is about where user attention stops. If a summarized answer satisfies the query, fewer users may click out to the pages that supplied the reporting, reviews or expertise behind it. Google’s blog frames the feature as a way to help users find “authentic voices” and “original content,” but that still leaves publishers dependent on how prominently Google surfaces those links. (support.google.com) ### How serious is the manipulation problem? The BBC reported this week that Google has been responding to efforts by companies and marketers to influence AI-generated answers. Google then updated its spam rules to say that attempts to manipulate generative AI responses in Search, including AI Overviews and AI Mode, violate policy. (blog.google) That change is notable because it treats AI-answer gaming as a search-integrity issue, not just a model-quality issue. Google has not said the problem is solved; its help pages still warn that AI responses may include mistakes, and the product remains labeled experimental in Labs access materials. (blog.google) ### What does Chrome have to do with this? CNET reported on May 22 that some Chrome users found a roughly 4GB Gemini Nano model had been downloaded onto their devices without clear notice. The report tied the download to Google’s push toward more on-device AI features in Chrome. (vuink.com) Google separately said last month that AI Mode in Chrome would let users ask follow-up questions while browsing pages side-by-side. Taken together, the Search and Chrome updates show Google spreading Gemini across both the search box and the browser window. ### What happens next? (support.google.com) Google said it will keep refining AI Mode, expand its functions and update the interface over time. For now, the company is steering early users to Search Labs and continuing to add AI features across Search and Chrome, while publishers, researchers and users test how often the new answers send traffic outward — or keep it inside Google. (support.google.com) (blog.google) (cnet.com)