Google on May 19, 2026 announced a sweeping set of changes to Search at I/O 2026, replacing its default AI model, introducing persistent background agents, redesigning the search box for the first time in more than 25 years, and expanding personal data integration to nearly 200 countries - all without a subscription requirement for the majority of features.

Google's AI Mode has, according to the company, surpassed one billion monthly users in the year since its debut, with queries more than doubling every quarter since launch. That backdrop set the stage for what Elizabeth Reid, VP of Search, described as "the next step in our journey to bring together the best of a search engine with the best of AI." The I/O 2026 keynote, held on May 19, 2026, covered five distinct product areas: a new default AI model, a redesigned search interface, search agents, agentic coding, and the broader rollout of Personal Intelligence.

Gemini 3.5 Flash becomes the default for AI Mode

The most immediate change is a model upgrade. As of May 19, Google replaced the previous default in AI Mode with Gemini 3.5 Flash, described by the company as "our newest Flash model delivering sustained frontier performance for agents and coding." The swap applies globally, covering all users in every country where AI Mode is currently available.

AI Mode itself has a trajectory worth recalling. It first appeared in Search Labs in March 2025 as an experimental feature for Google One AI Premium subscribers, relying at that point on a custom version of Gemini 2.0. Google opened it to all US users without a waitlist on May 20, 2025, then expanded it internationally in October 2025 to more than 40 countries and 35 new languages. In July 2025, Google reported a 65% surge in visual searches as multimodal adoption accelerated alongside AI Mode rollout. The Gemini 3.5 Flash upgrade announced on May 19 continues that cadence of model improvements, though Google has not provided technical benchmarks or latency comparisons in its public announcement.

0:00
/0:34

A redesigned search box after 25 years

Alongside the model change, Google on May 19 introduced what it calls the biggest upgrade to its Search box in over 25 years. The new interface is described as "completely reimagined with AI," and the changes are functional rather than purely cosmetic. The box now dynamically expands to accommodate longer queries - a relevant design decision given that AI Mode queries are typically two to three times longer than conventional search inputs, a pattern documented consistently since the feature launched.

The redesigned box includes AI-powered suggestions that go beyond traditional autocomplete. Rather than predicting the next word in a short keyword string, the system is designed to help users articulate more complete questions. According to Google, the box also accepts text, images, files, videos, and Chrome tabs as inputs - consolidating multimodal search entry into a single interface element. The rollout began on May 19, in all countries and languages where AI Mode is currently available.

Google also made conversational continuity easier. Follow-up questions can now be asked directly from within an AI Overview, flowing into a conversational back-and-forth in AI Mode. Context carries across turns, and according to the announcement, the links and supporting articles surfaced become more relevant as a conversation deepens. This seamless experience went live on May 19 across desktop and mobile, worldwide. It connects to a pattern PPC Land has tracked since December 2025, when Google began testing a direct transition from AI Overviews into AI Mode on mobile devices via an "Ask Anything" button at the bottom of expanded results.

Search agents: information monitoring in the background

The most structurally novel announcement is the introduction of search agents - persistent, autonomous agents that operate within Search without requiring a user to actively submit queries.

The first category is information agents. These run continuously in the background, monitoring the web for changes relevant to a user-specified question. According to Google, an information agent will look across "everything on the web, like blogs, news sites and social posts, plus our freshest data, such as real-time info on finance, shopping and sports," and send an intelligent synthesized update when something relevant occurs. The update includes the ability to take action based on the finding.

The examples Google provided illustrate the scope of what these agents are intended to handle. A user hunting for an apartment can describe precise requirements - location, price, layout, proximity to public transport - and the agent continuously scans listings, notifying the user when a match appears. A user who wants to know when a specific athlete announces a product collaboration can receive an alert when a new release is confirmed. Neither case requires the user to return to Search and reformulate a query.

Information agents will launch first for Google AI Pro and Ultra subscribers this summer. No pricing for those tiers was included in the May 19 announcement.

The agent announcement extends the agentic direction Google has been developing in Search since at least mid-2025. In July 2025, Google introduced automated calling capabilities in the US, where AI Mode contacts local businesses on behalf of users to check availability and pricing. In November 2025, agentic booking for restaurants, events, and local appointments expanded across the United States, drawing on platforms including OpenTable, Resy, Tock, Ticketmaster, StubHub, SeatGeek, Booksy, Fresha, and Vagaro. In April 2026, agentic restaurant booking reached the United Kingdom with eight local partner platforms. The information agents announced on May 19 operate at a different layer - monitoring and alerting rather than transacting - but they sit within the same architectural direction that Nick Fox, Google's SVP for Search and Maps, described in detail in April 2026.

Agentic booking expands to local experiences and services

Separate from information agents, Google on May 19 announced an expansion of agentic booking capabilities within Search to a wider range of tasks, including local experiences and services. The existing framework - which routes users to partner booking platforms rather than completing transactions inside Google's interface - is now extending to categories such as home repair, beauty, and pet care. For those select categories, according to the announcement, users can ask Google to call businesses on their behalf.

0:00
/0:41

The booking expansion is described as rolling out to everyone in the United States this summer. The approach follows the same structural model as the November 2025 agentic features: Google acts as an aggregator and routing layer, with the final booking occurring through the partner's infrastructure. Pricing and availability information is described as surfaced in real-time, with direct links to complete the transaction through the provider of the user's choice. The karaoke example in Google's announcement - "finding a private karaoke room for six on a Friday night that serves food late" - illustrates the specificity of query the system is designed to parse.

Agentic coding in Search: generative UI and mini apps

Google also brought what it calls agentic coding capabilities into Search, powered by a combination of Gemini 3.5 Flash and a technology the company refers to as Google Antigravity. The result is that Search can build custom interfaces - described as "generative UI" - in response to queries, in real time.

0:00
/1:00

The practical scope includes visual tools and simulations tailored to specific questions. A user asking about astrophysics might receive a custom interactive visualization rather than a list of links or a text summary. A user asking how a mechanical watch works might see a rendered diagram assembled on the fly. According to Google, generative UI capabilities will be available for everyone in Search this summer, free of charge.

Beyond one-off answers, the system is also designed to handle recurring needs. Google describes the ability to build custom dashboards or trackers for ongoing tasks - wedding planning, a home move, a fitness routine. These are characterized in the announcement as "mini apps for your own specific tasks," coded by Search on demand, drawing on real-time sources including reviews, live maps, and local data like weather. The fitness tracker example given involves Search coding a custom tool that integrates with fresh sources and updates week over week.

The ability to build these custom experiences will be available in the coming months, starting first for Google AI Pro and Ultra subscribers in the United States.

Personal Intelligence expands globally without a subscription

The final major announcement from May 19 is the geographic expansion of Personal Intelligence in AI Mode. Starting May 19, Personal Intelligence is available to users in nearly 200 countries and territories across 98 languages, and - critically - without requiring a subscription.

Personal Intelligence allows users to connect third-party applications to AI Mode. Currently supported connections include Gmail and Google Photos, with Google Calendar listed as coming soon. When connected, the AI can reference personal data from those sources when composing responses: email history, photo metadata and content, and eventually calendar events. According to Google, the feature was "designed with transparency, choice and control at its core," and users choose whether to connect each application.

The data handling details carry practical weight. PPC Land's February 2026 coverage of Personal Intelligencedocumented that Google added source icons to AI Mode showing when responses draw from Gmail, Photos, YouTube, or Search data, and that disconnecting an app from Google's interface does not automatically delete the associated data from Gemini Apps Activity. Users who want to remove that data must separately delete Gemini Apps activity.

The subscription-free expansion represents a significant shift from the feature's earlier availability. Personal Intelligence initially launched for Gemini app subscribers in January 2026. Opening it to the broader AI Mode user base - which itself surpassed one billion monthly users according to the May 19 announcement - substantially increases the number of people whose personal data from connected apps will inform search responses. How Google handles data governance at that scale has not been specified in the announcement beyond the general transparency framing.

Context for the marketing community

For digital marketing professionals, the I/O 2026 Search announcements carry several distinct implications.

The model upgrade to Gemini 3.5 Flash affects response generation quality and speed in AI Mode. PPC Land has tracked since April 2026 how AI Mode in Chrome opens publisher links side-by-side rather than replacing the AI interface - a structural change that affects click attribution, session behavior, and how publisher pages load within a split-screen context. The Gemini 3.5 model operating behind that interface shapes the responses that determine whether users click through at all.

Information agents represent a qualitatively different challenge. An agent that monitors the web and surfaces synthesized alerts operates largely outside the traditional search funnel. A user notified by an agent about a product drop or a listing that matches their criteria may convert through a direct link in that alert rather than through any search results page. The attribution frameworks and conversion tracking that marketing teams rely on are not designed for this interaction pattern.

The agentic booking expansion to home repair, beauty, and pet care categories brings additional service verticals into a flow where Google intermediates between user intent and final transaction. The booking itself occurs through partner platforms, but visibility in those results depends on factors - including how businesses are represented in Google's data ecosystem - that differ from standard paid or organic search visibility.

SISTRIX's March 2026 data documented click-through rates at position one in Google Search collapsing from 27% to 11%, driven by AI feature expansion. The May 19 announcements add new surfaces - generative UI, mini apps, background agents, and expanded conversational continuity - that deepen that dynamic. Last quarter, according to Google, queries reached an all-time high. The question the marketing community has been grappling with for more than a year is whether that query growth translates into equivalent traffic to publishers and advertisers, or whether an increasing share of queries resolve inside Google's interface entirely.

Timeline

Summary

Who - Google, through Elizabeth Reid, VP of Search, made the announcements on May 19, 2026 at Google I/O 2026. The features affect AI Mode users globally, with new agentic capabilities targeting US users in the first instance and subscriber-only tiers (AI Pro and Ultra) for information agents and mini apps.

What - Google upgraded AI Mode with Gemini 3.5 Flash as the new default model; redesigned the Search box with dynamic expansion and multimodal input; introduced information agents that monitor the web continuously and send synthesized alerts; expanded agentic booking to home repair, beauty, and pet care in the US; launched agentic coding capabilities that generate custom UI and mini apps on demand; and extended Personal Intelligence - which connects Gmail and Google Photos to search responses - to nearly 200 countries and 98 languages, free of charge.

When - The announcements were made on May 19, 2026, at Google I/O 2026. The Gemini 3.5 Flash model upgrade, the redesigned Search box, and Personal Intelligence expansion began rolling out on May 19. Information agents, agentic booking expansion to new categories, and generative UI capabilities are expected to roll out in the US this summer, with AI Pro and Ultra subscribers receiving some features first.

Where - The redesigned Search box and Gemini 3.5 Flash model are available globally in all countries where AI Mode operates. Personal Intelligence is expanding to nearly 200 countries and territories across 98 languages. Agentic booking expansion and information agents are initially US-only, with broader rollout not yet specified.

Why - Google's stated goal is to extend Search beyond reactive query-and-response into a system that acts continuously on users' behalf. The one billion monthly AI Mode user figure and the reported all-time high in query volume last quarter provide the commercial rationale for deepening AI integration. For the advertising and publishing ecosystem, the practical question raised by each of these features is how user journeys - and the attribution of value along those journeys - change when Search operates as an agent rather than a retrieval index.

Share this article
The link has been copied!