Google Search Agents: How permanent AI agents are transforming search and what this means for SEO and GEO

09.08.2026
|
Dimitrij Tonet
|
9 min

Since mid-June 2026, Google Search has continued to work even when no one is actively searching. With the rollout of "Information Agents," Google has introduced a feature that monitors the web around the clock in the background, notifying users whenever something relevant happens. Announced at I/O on May 19, 2026, the feature was launched on June 12 for Google AI Ultra subscribers in all languages and markets where the AI mode is available. Switzerland has been included since October 2025, with support for German, French, and Italian.

While this may sound like a convenience feature for end users, it represents the most profound change for business visibility since the launch of AI Overviews. In this article, you will learn how Search Agents function technically, why this shifts the goal of optimization, and which three levers you should start pulling now.

Key takeaways

Google has rolled out persistent search agents. Information Agents continuously monitor blogs, news sites, social media posts, and real-time data from finance, shopping, and sports, alerting users as soon as something changes. They are triggered in AI mode with phrases like "keep me updated on" or "notify me when."

Search is shifting from a momentary action to a continuous state.
Previously, the goal was to be visible at the exact moment of a search query. Now, monitoring runs permanently in the background, and the critical moment is no longer predictable. Visibility has become a state you must maintain, rather than a result you achieve once.

Recency is becoming the most critical ranking factor for AI citations.
Data from Amsive in 2026 shows that about half of all citations in AI search systems come from content less than 13 weeks old. An Ahrefs analysis of nearly 17 million cited URLs paints the same picture: AI-cited pages are, on average, significantly fresher than traditional organic search results.

Cosmetic updates do more harm than good.
Crawlers compare page versions. Changing only the timestamp without making substantial content updates devalues your signal. Google explicitly recommends omitting the "lastmod" tag if the value cannot be maintained reliably.

Presence outside your own domain is gaining importance.
Agents explicitly search through trade media, news, and social posts. Your own website is now just one of several sources that shape your brand's image.

Action is required now.
The expansion to Google AI Pro is scheduled for this summer. Those who wait until then to bring their most important pages into a maintainable state will lose their lead time.

What Information Agents actually are

Google groups this new feature under the umbrella term "Search Agents." Information Agents are the first type. The user defines their interests once in AI mode, and from that point on, the agent takes over the monitoring.

Google provides two examples: Someone looking for an apartment can describe all their requirements at once and receive a notification as soon as a matching listing appears. Someone wanting to know when a specific athlete announces a new partnership gets the update without any further effort. The result in each case is a synthesized summary with links and the option to take immediate action.

Robby Stein, VP of Google Search, described the June 12 rollout by noting that search agents are now available to Ultra subscribers in all AI-mode languages and markets, with more user groups to follow in the summer. Technically, this is the logical continuation of what began with "query fan-out." Even in AI mode, Google breaks a query down into multiple sub-questions, searches in parallel, and compiles the answer from many sources. The agent turns this into a standing order.

In parallel, Google showcased other building blocks at I/O that point in the same direction. The new search bar is considered the biggest overhaul in over 25 years, accepting text, images, files, videos, and even open Chrome tabs. "Generative UI" builds relevant calculators, maps, or comparison tools directly into the response. Furthermore, agentic booking functions are being expanded to local services, including—in selected categories—calls that Google makes to businesses on the user's behalf.

Why this changes more than AI Overviews

AI Overviews and AI mode have changed what an answer looks like. Search agents are changing when a search even happens. This is a difference in category, not just degree.

Level Classical Search AI Mode Search Agents
Trigger: User types a query User asks a question in a dialogue System continuously searches in the background
Timing: Predictable, based on search volume Predictable, but less tied to keywords Unpredictable, can occur at any time
Optimization goal: Rank in positions 1 to 10 Be cited in the generated answer Be up to date and discoverable at the moment of change
Most important signal: Relevance and backlinks Extractability and authority Freshness, consistency and machine readability
Source base: Indexed websites Websites plus real-time data Websites, specialist media, social posts and real-time data
User interaction Click on a search result Answer with a source reference Push notification with a summary

The practical consequence can be illustrated with an example. Previously, an interior designer in Zurich was found when someone actively searched for one. In the future, a potential customer might simply state that they are looking for a company in Zurich for kitchen renovations within a specific price range and with availability in the autumn. The agent then monitors the situation over several weeks. Whether this business appears in the push notification is not decided on the day of the search, but rather on whether its information is current, complete, and machine-readable at any given moment. This is precisely where the work of an SEO agency shifts from one-time optimization to the ongoing maintenance of key content.

What this means for Swiss companies in concrete terms

Switzerland is not being affected by this development later than other markets, but simultaneously. AI mode has been available here in three national languages since October 7, 2025, and agents have been explicitly enabled in all AI mode languages. Anyone waiting for Google to roll out later in Switzerland no longer has that buffer.

Three points are particularly relevant for Swiss companies.

Multilingualism is becoming a consistency issue. An agent monitoring in French will draw on French sources. If your German site is well-maintained but your French site has had outdated prices and opening hours for two years, it creates a contradictory image of your brand. For AI systems, contradiction is a trust issue, not a translation issue.

Local service providers are moving to the center. Google is specifically expanding agentic booking functions to include local experiences and services. Availability, prices, and service descriptions are thus becoming ranking-relevant data. A business whose website shows a 2024 price list will simply be skipped by an agent.

SMEs have a speed advantage here. Large organizations take months to change approval processes for content updates. A Swiss SME can bring its twenty most important pages into a clean state in just a few weeks. In our daily work as a Swiss SEO agency we regularly see that smaller companies catch up faster in AI visibility than in classic rankings.

Lever 1: Recency is becoming a hard signal

When an agent is constantly looking for changes, recency is no longer a soft quality factor, but a condition for entry. The data on this is now robust.

In 2026, Amsive found that about half of all citations in AI search systems come from content less than 13 weeks old. A study by Ahrefs based on nearly 17 million cited URLs points in the same direction. Analyses of the half-life of content in AI answers describe a steep decline in the first year, which then transitions into a long tail.

The distinction between genuine and artificial recency is important here. Crawlers compare page versions with each other. If the timestamp changes but the content does not, the signal is devalued. Google is unequivocal about lastmod and recommends omitting the value entirely rather than maintaining it inaccurately. This is not an anti-freshness stance, but an anti-fabrication stance.

Relevance has three layers that must align.

Layer What it means Where it belongs technically Typical mistake
Content freshness: Figures, examples, screenshots and recommendations reflect the current reality In the body text and tables on the page The year in the title is updated, but the content remains from 2024
Visible freshness: The user can immediately see when the content was last reviewed A visible "Last updated" line near the top of the page Only the original publication date is displayed
Machine-readable freshness: Crawlers and LLMs can detect the change in a structured way Schema.org dateModified, sitemap lastmod, HTTP header Last-Modified The three values contradict each other

A practical addition that costs little but delivers much is a brief changelog on your most important pages. Two or three lines noting what was updated are enough. For users, this provides transparency; for machines, it is verifiable proof that the site is being maintained.

Lever 2: Presence beyond your own website

Google explicitly states that information agents monitor blogs, news sites, and social posts. Your own domain is therefore just one source among many. This observation aligns with what already applies to generative search engines, where the majority of brand mentions for discovery queries come from third-party sources, with only a small fraction originating from the brand's own domain.

In practice, this means that off-page work is no longer primarily about link building, but about mention building. Industry articles, entries in relevant directories, maintained review profiles, contributions to professional forums, and regular activity on LinkedIn now serve a different function than before. They are not just authority signals, but additional observation surfaces for agents running in the background.

Consistency across all these platforms is crucial. If your website names a service differently than your LinkedIn profile, and an industry directory uses a third variation, your entity fragments into several blurred versions for the model. This is exactly where the work of a GEO agency comes in, by standardizing entities, terminology, and core messaging across all channels.

Google itself highlighted how closely SEO and GEO are linked in its official guide to AI visibility from May 15, 2026 . According to the guide, GEO and AEO are not separate disciplines, but rather SEO fundamentals applied to AI interfaces. If your foundation isn't solid, GEO measures will have little effect.

Lever 3: Machine readability and efficiency

An agent that runs continuously processes enormous amounts of pages. It will not patiently wait for JavaScript to load on every page, nor will it read through 800 words of introduction to find a price. Efficiency thus becomes a technical criterion.

Four points are most important here.

  • Server-side rendering for everything relevant. Many AI bots render JavaScript only to a limited extent or not at all. Prices, availability, opening hours, and core statements must be present in the initial HTML.
  • Answer before context. Addy Osmani, Director of Engineering at Google Cloud AI, argued in April 2026 that token efficiency is a genuine optimization factor for agent-readable content. Pages that bury their core message deep within the body text are more frequently skipped or inaccurately summarized.
  • Structured data that reflects reality. Product, LocalBusiness, FAQPage, Article, and Offer are not just decoration. For an agent, they are often the only reliable source of data.
  • Direct delivery where it counts. For companies with frequently changing data, delivering content via APIs and feeds can be the difference between being found and being overlooked.

One side effect of the I/O announcements deserves special attention. Google is bringing SynthID and C2PA content credential verification to Search and Chrome. This allows AI-generated content to be detected and labeled at scale. Anyone publishing mass-produced, unedited AI text in the second half of 2026 is taking a risk that didn't exist in this form before.

Your first 30 days

You don't have to rebuild your entire website. This sequence will have the biggest impact in one month.

Week 1: Inventory. Define the twenty to fifty pages that truly matter for purchasing decisions. For each page, check whether the figures, prices, examples, and contact details are still accurate. Note which pages are outdated in terms of content, regardless of what the date says.

Week 2: Synchronize signals. Align the visible date, the dateModified in your schema markup, and the lastmod in your sitemap for these pages. If you cannot reliably maintain the lastmod, it is better to remove it. Additionally, check whether your server is delivering a correct Last-Modified header.

Week 3: Update substance. Truly update the content-outdated pages with new figures, current sources, corrected recommendations, and at least one new content element. Keep the URL the same so that the authority you have built is preserved.

Week 4: Check rendering and consistency. Disable JavaScript in your browser and visit your most important pages. Is the core information still visible? At the same time, compare the information on your website with your LinkedIn profile, your directory listings, and your foreign-language pages, and resolve any contradictions.

After that, you should establish a fixed rhythm. Pages with volatile information like prices or availability need a monthly review, while guide content needs a quarterly one. This ongoing maintenance is the part that many companies underestimate, and therefore a reason to consider working with a professional GEO agency .

FAQ on Google Search Agents

What are Google Information Agents?

Information Agents are AI agents in Google Search that continuously monitor the web in the background for a topic defined by the user and send a summarized update with links whenever relevant changes occur. They are the first available type within the broader Search Agents category.

When did Search Agents become available?

Google announced the feature at I/O on May 19, 2026, and made it available to Google AI Ultra subscribers on June 12, 2026, across all languages and markets where AI Mode is available. Expansion to Google AI Pro was announced for summer 2026.

Are Information Agents available in Switzerland?

Yes. AI Mode has been available in Switzerland in German, French, and Italian since October 7, 2025, and the agents were launched across all AI Mode languages and markets. An appropriate Google subscription is required.

How can I optimize my website for Search Agents?

Three measures are particularly effective. Keep your most important pages up to date and support this with consistent date signals. Make sure key information is available directly in the HTML without relying on JavaScript and is properly marked up with structured data. And build visibility beyond your own domain, as agents explicitly evaluate industry publications, news sources, and social media posts as well.

Is it enough to regularly update the modification date?

No, and it can even be counterproductive. Crawlers compare different versions of a page. A changed timestamp without a substantial change to the content may be treated as an unreliable signal. Google recommends omitting lastmod if the value cannot be maintained reliably.

Does this make GEO more important than traditional SEO?

The two cannot be meaningfully separated. In May 2026, Google itself clarified that GEO and AEO are not separate disciplines, but rather solid SEO fundamentals applied to AI-driven search experiences. Without sound technical foundations, clear structure, and reliable content, GEO measures have limited impact.

Conclusion: Visibility is no longer a result, but a state of being

Search Agents mark the point where search stops being a one-off event. Previously, visibility was something you achieved and then defended. But when an agent monitors in the background for weeks, visibility is decided at a moment you cannot know or plan for. The only possible response is to be in a state of constant readiness.

This is less dramatic than it sounds, and it rewards the very companies that already operate with integrity. If you maintain your pricing, update your technical content, provide structured data, and speak with a consistent voice across all channels, you will meet the requirements of Search Agents almost automatically. Conversely, those who run a single campaign once a year and do nothing in between will find themselves increasingly absent in a world of permanent, observational search.

The good news is that the fundamentals remain unchanged. The principles are the same as they have always been; it is just that the tolerance for neglect has diminished.

/ Contact us

Are you ready to dominate the first page on Google?

Our team of SEO experts is ready to help you. Discover your SEO potential in a free initial consultation!

Oops! Something went wrong while submitting the form.