Google has begun deploying autonomous "information agents" inside its AI Mode search interface, a feature that monitors the web continuously on behalf of users and delivers updates only when something actually changes — representing a structural break from the one-query-at-a-time model that has defined search for three decades.
The feature, unveiled at Google's I/O 2026 developer conference as "Search agents," started rolling out on June 12 to Google AI Ultra subscribers. The first variant available, called the information agent, operates in the background 24 hours a day, scanning blogs, news sites, social media, financial data feeds, and real-time information streams tied to whatever topic a user specifies.
Users activate an agent by phrasing a prompt with either "keep me updated on" or "alert me when" followed by a subject. Google's documented example: a user inputs apartment requirements, and the agent sends a notification when a new listing appears that matches. The agent delivers a summarized update rather than a raw feed of results — the platform estimates that a user who would otherwise check multiple times a day for a new apartment listing is replaced by a single system that pings them when something material changes.
The differentiation from existing Google notification tools is material. Gemini's scheduled actions currently check at most once per day, and Gemini Spark uses a fixed 15-minute polling interval regardless of whether new information exists. Information agents remove the interval constraint; they trigger only when the system determines an update is relevant. That shift from time-based polling to event-based delivery is what separates a handy tool from a genuinely useful agentic interface.
The launch window is narrow for now. AI Mode information agents are available across all supported languages and markets for eligible AI Ultra subscribers, who pay $99.99 per month or $199.99 per month for the highest tier. Google AI Pro subscribers are expected to gain access over the summer of 2026. The company has not yet announced plans for a free tier of the agent functionality, which means the feature is starting as a premium product aimed at power users — exactly the cohort that will generate the most rigorous feedback.
The rollout lands at a moment of high scrutiny over AI in search. In June 2026, industry reports confirmed that bots now generate more web traffic than human users for the first time, raising questions about how search engines index, prioritize, and present information in an environment where AI agents are both consumers and producers of web content. Google's agent infrastructure is simultaneously an answer to that problem — users receive synthesized summaries instead of querying search repeatedly — and an accelerant of it, since each active agent generates additional automated requests against news and data sources.
Google is trying to thread a narrow needle: aggressively introducing AI products that replace traditional search behaviors while protecting the advertising revenue built on decades of human-initiated queries. The information agent does not remove the search box; it adds a layer above it. Whether users adopt agents at scale will have as much to do with trust and habit as with technical capability — trust that the agent is not missing something important, and habit enough to configure it in the first place.
Image source: v3b.fal.media