Google AI Search, shorthand for Google Search rebuilt with artificial intelligence (AI, software that can generate and act on information), is moving from a results page toward a task layer: Gemini 3.5 Flash now powers AI Mode, the search box is being remade for longer prompts, and agents can monitor web changes for users.
Publisher traffic is the first casualty many readers will notice. Google is also moving the valuable moment of intent closer to alerts, comparisons, bookings and checkout, where structured data and transaction access matter more than a top organic rank.
An AI Search Box Rewrites the First Click
Google, the Alphabet-owned technology company, used its I/O developer conference on May 19 to give Search a new front door. In Google’s I/O Search announcement, Elizabeth Reid, vice president of Search at Google, said AI Mode had passed one billion monthly users and that the company was making Gemini 3.5 Flash the default model for AI Mode globally.
The visible change is the box itself. Google says the new Search box expands so users can describe more complex needs, then offers AI-powered suggestions that go beyond autocomplete. The input menu also widens: text, images, files, videos and Chrome tabs can become search material. That matters because the classic query was a short instruction. The new query can be a messy bundle of context.
Follow-up questions now sit closer to the main results experience as well. A user can move from an AI Overview into a conversation with AI Mode without starting over. Google AI Search is now a task layer, not just a faster way to summarize pages. The first click becomes less about choosing a link and more about letting Google decide which sources, formats and next actions fit the request.
Scale Turns the Feature Into Infrastructure
Sundar Pichai, chief executive of Alphabet and Google, put the scale plainly in his I/O opening keynote. AI Overviews now reach more than 2.5 billion monthly active users, AI Mode has passed one billion monthly active users, and the Gemini app has surpassed 900 million monthly active users. Those are not lab metrics. They are distribution.
- 2.5 billion monthly active users for AI Overviews, according to Pichai.
- 1 billion monthly active users for AI Mode, reached within about a year.
- 900 million monthly active users for the Gemini app, up from 400 million at the prior I/O.
- $180 billion to $190 billion in expected capital expenditures this year, about six times the 2022 level Pichai cited.
That spending figure explains why Search is central to the story. Alphabet can justify enormous computing costs if the output protects the most valuable habit on the consumer internet. The model upgrade is not only about better answers. It is about turning a search session into a longer, more instrumented interaction that can support shopping, finance, travel, local services and software tasks.
Publisher Economics Face the First Stress Test
The web’s old exchange was simple enough: publishers allowed search crawlers to index their work, and search engines sent back readers. Cloudflare, the internet infrastructure company, says that exchange is weakening. In Cloudflare’s crawl and referral analysis, the company said AI platforms can make vastly more page requests than referrals, with Anthropic’s Claude measured at roughly 70,900 HTML page requests for every HTML page referral during one June week.
Google’s position is more complicated because its crawler still supports both classic search and AI features. Cloudflare’s data also showed that referral traffic in its measured set was still dominated by Google, while AI platform referrals remained much smaller. That makes Google’s upgrade uniquely sensitive: publishers cannot treat Google like any other AI crawler without risking their place in search discovery.
BrightEdge, an enterprise search engine optimization (SEO, the practice of improving visibility in search results) platform, sees a different version of the same squeeze. In BrightEdge’s AI Overviews study, the firm said content impressions rose more than 49 percent after the launch of AI Overviews while click-throughs fell nearly 30 percent. More exposure, fewer visits. That is a harsh trade for ad-supported sites.
Penske Media, the publisher behind Rolling Stone, Billboard and The Hollywood Reporter, turned the economic argument into a legal one. In the Penske Media complaint against Google, the company alleges that AI Overviews and AI Mode use publisher content in ways that reduce the need to click through. Google has argued in its own materials that AI features include links and can send higher quality visits. The dispute now rests on a metric publishers care about most: paid attention.
Transaction Owners Gain New Leverage
The strongest position in the new Search may belong to companies that control fresh data or the final action. A ranked article can inform an answer. A merchant feed, booking system, map listing, inventory table or payment path can complete the job. That changes who has leverage when Google turns a query into a task.
| Stakeholder | Why Its Position Improves | Why It Still Hurts |
|---|---|---|
| Owns the interface, user context and ad matching layer. | Faces higher compute costs and more antitrust scrutiny. | |
| Merchants and travel platforms | Can plug offers, prices and booking paths into AI-assisted decisions. | Depend more on Google’s display choices and data rules. |
| Publishers | Strong reporting can still feed citations and authority signals. | Summaries may satisfy readers before a site visit. |
| SEO agencies | Demand rises for AI visibility, brand monitoring and structured data work. | Old rank reports lose power when answers vary by conversation. |
| Local service businesses | Agents can surface availability, pricing and contact options directly. | Customer relationships may begin inside Google’s interface. |
The table’s lesson is blunt: clicks are no longer the whole market. In a blue-link world, visibility was measured by rank and visit volume. In an agentic search world, visibility also means being legible to machines, trusted by the model, connected to live data and ready for a user who may never browse the site in the old way.
Ads Move From the Margin Into the Answer
Google’s ad team is not treating AI Search as a side experiment. In Google’s AI-era Search ads post, the company said it is testing Gemini-built ad formats, including Conversational Discovery ads and Highlighted Answers. It also said Direct Offers will surface in AI Mode responses, with native checkout integration for Universal Commerce Protocol (UCP, a merchant standard for completing purchases) participants and coming travel offers from partners such as Booking and Expedia.
That places advertising closer to the answer, not just above or beside it. Google Ads help pages say ads in AI Overviews can be matched using the user’s query and the content of the AI Overview, which lets Google infer commercial intent even when the original query does not look like a purchase query.
- Retailers need clean product feeds, current prices and clear return policies, not just keyword lists.
- Travel companies need inventory and offer data that can survive comparison inside a generated answer.
- Lead-generation advertisers need to measure quality after the click, because AI may prequalify or misread intent.
- Brands need tighter negative guidance so automated matching does not place them in awkward contexts.
This is the part of the shift that should make advertisers both interested and nervous. Better intent matching can raise conversion quality. Less visibility into where and how a generated answer shaped the buyer can make campaign diagnosis harder.
Regulators Now Have an AI Version of Search
The U.S. Department of Justice, the federal antitrust enforcer, already has Google under remedies tied to search distribution. In an April 2026 remedies statement on Google Search, the department said the court barred certain exclusive contracts involving Google Search, Chrome, Google Assistant and the Gemini app, and ordered Google to make some search index and user-interaction data available to rivals.
Those remedies show why AI Search is not a normal product redesign. The interface that regulators spent years studying is changing while the case law is still catching up. If search becomes a blend of answer, agent, ad inventory, checkout path and personal context, the old question of who gets the default search box becomes only one part of the market.
The European Commission, the European Union’s executive arm, opened its own formal investigation in December into Google’s use of online content for AI purposes. The official Commission probe into online content for AI focuses on whether Google used publisher and YouTube content on unfair terms, including for AI Overviews and AI Mode.
Regulators now have to define harm in a product where the user may be satisfied, the advertiser may pay, the publisher may lose the visit, and Google may still show links. That is a messy case to prove. It is also the market Google is building in public.
The Web Bargain Depends on Better Signals
For site owners, the near-term answer is not to abandon Google. It is to stop treating search visibility as a single ranking problem. Pages still need authority and usefulness, but machine-readable structure, fresh feeds, clear authorship, licensing choices and direct audience channels now carry more weight.
Cloudflare, publishers and SEO firms are all circling the same new bargaining table: who gets to use the web’s work, what payment or traffic flows back, and which technical controls are strong enough to matter. Google’s advantage is that users already begin there. Publishers’ advantage is that Search still needs credible information from outside Google to stay useful.
If AI Search sends qualified users back to original sources, the open web bends but keeps producing. If it absorbs the answer, the comparison and the transaction without enough value flowing back, the web’s best work will move behind gates that even Google cannot summarize for free.
