Google AI search announcements 2026: I/O and AEO

Google I/O 2026 ran on May 19 and 20. Most of the press attention went to Gemini 3.5 Flash, Universal Cart, Antigravity, smart glasses, and a personal AI assistant called Gemini Spark. But the parts that matter most for anyone running a website did not get the same airtime, and a few of them didn't even happen on the keynote stage.

Search itself got a redesign Google is calling its biggest in over 25 years. Google Search Central published a formal generative AI optimization guide four days before the keynote. Google Marketing Live, the week after, folded Gemini into nearly every advertising and merchant product. Read together, these updates change how to plan for Google visibility through the rest of 2026 and into 2027.

This piece walks through what was announced, what the data actually says about traffic, and what the most credible research suggests doing next.

What did Google announce for AI search in 2026?

In 2026, Google announced new AI search features in AI Mode and AI Overviews, more links to websites in AI answers, and new tools for site owners. Its written guidance for site owners came on May 15, 2026: a guide to optimizing for generative AI features. The table lists each one by date, oldest first.

Key takeaways

• The scale numbers Google shared: AI Mode hit 1 billion monthly users one year after launch. AI Overviews now reaches 2.5 billion monthly users. AI Mode queries have more than doubled every quarter since launch (Google Keyword blog, May 19, 2026).
• Search behavior is changing in measurable ways: The average AI Mode query is roughly three times longer than a traditional Search query, and more than one in six U.S. searches now uses voice or images (Google AI Mode U.S. insights).
• Google's official position on AEO/GEO: AI Overviews and AI Mode run on the same crawl, index, and ranking systems as classic Search. There is no separate AI index, no special schema, and no LLM-specific upload format that earns citations (Google Search Central, May 15, 2026). Snippet eligibility is the floor.
• The click data is real but contested: Pew Research found an 8% click rate on results with an AI Overview versus 15% without one (Pew). One SEO software vendor reports a 58% CTR drop for the top-ranking page on AI Overview queries (300,000 keywords, December 2025). Different studies measure different things and cannot be compared directly.
• Citation is the defense: One search agency's analysis of 3,119 queries across 42 organizations found that brands cited inside AI Overviews earned 35% more organic clicks and 91% more paid clicks than uncited brands.

‍

Search itself: what Google announced

Liz Reid, who leads Search at Google, summarized the change in the opening of the official blog post:

‍"We're bringing our advanced model capabilities to Search with new AI features, enabling you to use agents just by asking a question. We're also introducing a new, intelligent AI-powered Search box, marking its biggest upgrade in over 25 years."

Here are the Search announcements that matter for visibility:

1. AI Mode is now powered by Gemini 3.5 Flash as the default model worldwide. Google says Gemini 3.5 Flash runs roughly four times faster than competing frontier models on output tokens per second, while outperforming the prior Gemini 3.1 Pro on coding and agentic benchmarks like Terminal-Bench 2.1 (76.2%), GDPval-AA (1656 Elo), and MCP Atlas (83.6%).
2. A new intelligent Search box that expands dynamically as you type, accepts text, images, files, videos, or Chrome tabs as inputs, and offers AI-powered query suggestions. Reid described it the same way in press briefings: "This new search box puts our most powerful AI tools right at your fingertips, and you can ask across modalities with text, images, files, videos, and search reasons across them all."
3. A smooth flow between an AI Overview and AI Mode. You can ask a follow-up right from an AI Overview and the conversation carries context into AI Mode. This is live worldwide on desktop and mobile today.
4. Search agents, starting with information agents that monitor the web 24/7 for whatever you care about. Information agents roll out to AI Pro and Ultra subscribers this summer. Agentic booking expands to local services like home repair, beauty, and pet care, where Search can call businesses on a user's behalf in select U.S. categories this summer.
5. Agentic coding in Search, powered by Google Antigravity. Search now generates custom layouts on the fly, including visual simulations, interactive widgets, and mini-apps for ongoing tasks like wedding planning or fitness tracking. Generative UI is free for everyone in Search this summer.

Outside Search, Google introduced Universal Cart, a shared shopping cart that follows the user across Search, Gemini, YouTube, and Gmail. Per Google Shopping, it runs on a Shopping Graph of more than 60 billion product listings and on two new protocols that handle agent-to-merchant communication: Universal Commerce Protocol and AP2. The U.S. rollout begins this summer across Search and Gemini. Canada and Australia follow in the coming months.

‍

How people are searching now

The most useful data Google released came from a parallel blog post the same day as the keynote, How AI Mode is changing and expanding the way people search. The numbers describe how people actually use AI Mode in the U.S.

• More than one in six U.S. searches now uses voice or images.
• Image-based queries in AI Mode are growing more than 40% month-over-month since launch.
• The average AI Mode query is triple the length of a traditional Search query.
• Planning queries in AI Mode have grown 80% faster than AI Mode queries overall in the past six months.
• Brainstorming queries (phrasing like "where to," "ideas for," "what should I") have grown 30% faster than the overall AI Mode mix since launch.
• Follow-up queries in AI Mode have risen more than 40% per month on average in the U.S.

A two-word keyword is now competing for attention with a three-sentence prompt that already contains context, constraints, and an implied follow-up. A page optimized only for the keyword has fewer ways to win. Reid said it directly during the press briefing: "It's not just that people are searching more, it's that they're searching differently. They're fully expressing their questions in granular detail, asking those follow-up questions and searching across modalities."

‍

Google's quieter but bigger update: the May 15 documentation

Four days before the keynote, Google Search Central published Optimizing your website for generative AI features on Google Search. It is Google's first formal, on-record documentation specifically for visibility inside AI Overviews and AI Mode. A few quotes from the document are worth reading literally.

‍"From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO."
‍
"You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in generative AI search."

"There's no requirement to break your content into tiny pieces for AI to better understand it. Google systems are able to understand the nuance of multiple topics on a page."

"Just because a page meets all requirements, best practices, and complies with the policies, doesn't mean that Google will crawl, index, or serve its content."

Google's position: AI Overviews and AI Mode use the same crawl, index, and ranking systems as classic Search. There is no separate AI index, no special AEO schema, and no LLM-specific upload format that earns citations. Snippet eligibility is the prerequisite for appearing in any generative AI feature.

On the AEO/GEO label question, Google argues the work to win citations is still SEO. In practice, many teams find a separate working label useful because the content choices, sourcing standards, structure, and measurement now deserve their own focus. The label is less important than the work itself. Either way, the operational change is the same: build content AI systems can trust, summarize, and quote with first-hand experience, named experts, clean structure, real statistics, and accurate operational data.

Two themes the guide returns to repeatedly are what Google calls "valuable, non-commodity content" and a clear technical structure. The phrase non-commodity does a lot of work in the document. Google's own example contrasts a generic "7 Tips for First-Time Homebuyers" with "Why We Waived the Inspection and Saved Money: A Look Inside the Sewer Line." The second one is harder for an AI to summarize away because the experience is specific.

‍

Google Marketing Live: AI in the ad and merchant stack

A week after I/O, Google Marketing Live folded Gemini into nearly every paid and merchant product. The pieces that matter for organic teams:

• New Gemini-built ad formats for AI Mode and conversational Search. Per Marketing Dive, these formats can show "when relevant not just to the person's query, but also the AI Overview response."
• Direct Offers in AI Mode ads, where retailers upload discounts, giveaways, and coupons and let Gemini match or bundle the most relevant offer to the user's search.
• AI performance insights in Merchant Center, which let brands compare share of voice across AI Mode, AI Overviews, and Gemini app shopping journeys against similar brands. Rolling out in Australia, Canada, India, New Zealand, and the U.S.
• Conversational attributes in Merchant Center, which let retailers write product descriptions in the more natural language people use in AI search.
• Ask Advisor, a Gemini-powered conversational interface across Google Ads, Analytics, Marketing Platform, and Merchant Center.
• UCP-powered checkout expanding to Canada and Australia in the coming months, with the United Kingdom to follow.

Citation share is the proportion of all citations, across a set of questions, that point to your domain; share of voice applies the same idea to mentions, or to mentions and citations together, depending on who reports it.

The reason this belongs in an SEO conversation: paid and organic are now sharing real estate inside AI surfaces. The same Merchant Center data feeds visibility in both. Clean product data, accurate hours, structured business profile information, and clear pricing are the same inputs that influence AI Overviews placement. The line between merchandising and content is thinner than it was a year ago.

‍

What the traffic data really shows

The CTR story has gotten clearer in 2026, but the headline numbers need careful reading. Three studies do most of the work in this area.

Pew Research Center (July 2025) tracked the actual browsing behavior of 900 U.S. adults across roughly 68,879 Google searches in March 2025 (Pew Research Center).

• 58% of participants saw at least one AI Overview that month.
• Users clicked a traditional search result on 8% of visits with an AI Overview versus 15% without one. A 47% relative drop in click probability.
• Clicks on the links inside an AI Overview itself: about 1% of visits.
• Sessions ended after the AI Overview page on 26% of visits, versus 16% without one.

One SEO software vendor analyzed 300,000 keywords in December 2025 using Google Search Console data, comparing AI Overview keywords against informational keywords without one. AI Overviews correlated with a 58% lower CTR for the top-ranking page, up from 34.5% in the firm's April 2025 study. The 58% number refers to position-one CTR on AI Overview queries, not aggregate traffic loss across a whole site.

A search marketing agency analyzed 3,119 informational and educational queries across 42 client organizations from June 2024 to September 2025, covering 25.1 million organic impressions and 1.1 million paid impressions. Three numbers from the study get cited everywhere:

• Organic CTR fell 61% on queries with an AI Overview.
• Paid CTR fell 68% on the same queries.
• Brands cited inside the AI Overview earned 35% more organic clicks and 91% more paid clicks than brands not cited.

That last finding is the most actionable. Citation is the most reliable defense against the click squeeze, no matter which study you trust most. The agency's lead researcher added: "We cannot definitively prove that citation causes higher CTRs; it's equally possible that brands with stronger authority and higher baseline CTRs are simply more likely to be cited by Google's AI. What we can say with confidence is that queries where you're cited consistently outperform those where you're not, regardless of the causal direction."

Two more data points are worth knowing. The agency's Q1 2026 update reported that organic CTR on AIO-present queries rose from 1.3% in December 2025 to 2.4% in February 2026, though the agency characterizes this as stabilization rather than a recovery. Another SEO software vendor, looking at over 200,000 keywords before and after they began triggering AI Overviews, found that zero-click rate on those same keywords actually dropped slightly, from 33.75% to 31.53%, suggesting that AI Overviews often land on queries that were already low-click to begin with.

The honest read: average traffic to top-ranked pages is meaningfully lower when an AI Overview appears, the size of the impact varies a lot by query type and industry, and being cited inside the AI Overview correlates with the smallest losses, sometimes outright gains in clicks per impression.

‍

What to do about it: a practical playbook

Six priorities for the rest of 2026. None are new tricks. The change is in priority order and standard of execution.

1. Earn snippet eligibility before anything else. Google's May 15 guide is explicit: pages must be indexed and eligible for a normal snippet before they can appear in any generative AI feature. Core technical SEO (crawlability, valid metadata, page experience, internal linking, content quality) is still the floor.
2. Build pages an AI can actually quote. The Princeton GEO paper, the agency's citation data, and Google's "non-commodity" language all point the same way: original statistics, named experts, real first-hand experience, and properly attributed sources earn citations more reliably than generic restatements. A page that says "we tested X across 200 customers and found Y" can be cited. A page that says "X is important for many businesses" cannot.
3. Write for whole decisions, not single keywords. Planning queries in AI Mode are growing 80% faster than the overall AI Mode mix. A buyer comparing options on Sunday, narrowing them on Monday, and asking for a final recommendation on Tuesday is using one search surface for all three steps. Content that answers the entire decision arc (comparisons, tradeoffs, objections, proof) earns more impressions across that arc.
4. Add real images and video. Google's guide calls out "high quality images and video" as additional opportunities to appear in AI features. Multimodal AI Mode queries are growing the fastest of any query type. YouTube has become one of the most-cited domains in AI Overviews. Sites that publish only text are competing for fewer surfaces than they could be.
5. Tighten your operational data. Google now calls businesses on a user's behalf in select U.S. categories. Universal Cart watches for price drops, restocks, and product compatibility. Stale hours, wrong prices, missing categories, or messy product feeds become a visibility problem, not just a conversion problem. Merchant Center, Google Business Profile, and structured business data deserve the same care as the homepage.
6. Measure more than rank. A useful 2026 dashboard tracks three things in addition to rank: AI Overview presence on your priority queries, citation share inside those overviews, and branded demand. Merchant Center's new AI performance insights give a comparable share-of-voice number on AI surfaces. That metric did not exist a year ago and will matter more by year-end.

‍

What is settled

Visibility inside AI-generated answers now deserves its own strategy and its own measurement. Traffic patterns are clearly changing. The size of the impact is contested. The work that helps with both AI visibility and traditional rankings is mostly the same work, executed with a sharper eye on credibility, structure, and proof.

‍

What did Google I/O 2026 change for SEO and AEO?

If the only thing you took from the keynote was "Google announced more AI," the more useful read is this. In the same two-week window, Google tightened Search, Search Central guidance, ads, shopping, and merchant tooling around one principle. Classic SEO is still the base layer. On top of that base, the businesses that win in 2026 are the ones that are easy for AI systems to crawl, rank, trust, summarize, cite, and act on. That second set, trust, summarize, cite, and act on, is what the practical work of AEO looks like under any name. Our free AI search tools check the crawl part, and a free Lenz report shows whether AI engines cite you for your buyers' questions.

If you want to go deeper, the primary sources are linked inline throughout this post. The most useful starting points are Google's May 15 generative AI optimization guide and the AI Mode U.S. insights report. They cover the official position, the new behavior data, and the strongest academic evidence on what actually helps.

Written bySoroush MehriCo-Founder and Chief AI Officer

Ex-Microsoft. 10+ years of experience in AI/ML, MSc from Université de Montréal (Mila). US patent holder.

Reviewed byMojtaba EbadiCo-Founder and Chief Operating Officer

Sr. Engineering Leader with 15+ years of experience. Ex-ApplyBoard. AWS Certified Solutions Architect.