AI Mode alerts fire on what changed

An AI Mode information agent is a standing request inside Google Search that keeps checking the web for a user and sends a synthesized update, with links, when the thing they asked about changes.

Until last week this was a paid feature. On September 28, 2026, Robby Stein, Google's VP of Product for Search, said the monitoring capability Ultra and Pro subscribers had been using was rolling out "to everyone globally," as reported by Search Engine Roundtable. The examples Google offers are ordinary: new restaurants or pop-ups nearby, holiday activities for kids, a product coming back in stock, a price drop.

That changes what a citation in AI Mode is. A normal AI Mode answer is generated when someone asks. An alert is generated when the web changes. The question was asked weeks ago, and the page that gets credited is the one where Google could see something new.

#What Google has said about how it works

The feature was announced at I/O on May 19, 2026. Google's I/O Search post describes agents that run "in the background, 24/7" and scan blogs, news sites and social posts plus real-time finance, shopping and sports data. When the criteria are met, the user gets a synthesized update with links and an option to act on it.

The mechanics reported since then are consistent across outlets:

• Trigger. A user types "keep me updated on" or "alert me when" in AI Mode (9to5Google). Google also suggests monitors during normal searches, per Digital Information World.
• Delivery. Updates arrive as push notifications in the Google app, and active monitors live in AI Mode history where the user can refine or turn them off (TechCrunch). TechCrunch frames it as the next step after Google Alerts, which launched in 2003.
• Sources. Websites, forums, social posts, real-time data and the Shopping Graph, which Google puts at more than 60 billion products (Search Engine Roundtable).
• Scale of the host surface. Google says AI Mode passed one billion monthly users a year after launch (Google). Nobody has published how many of them have a monitor running.

What Google has not said matters as much. There is no public statement on how often an agent rechecks, how it decides a change is significant, or how it picks which sources go in the update. There is also no reporting. Search Console does not show alert impressions, and a push notification leaves no referrer you can segment unless the user taps through.

#Why the unit of competition shifts to the change

A one-off answer rewards the page that best explains a topic. A monitor rewards the page that first and most clearly shows a new state of the world: the restock, the new price, the opening date, the announcement. Those are different jobs, and a lot of sites are built only for the first one.

Consider where a typical business publishes change today. A price cut lives in a cart banner. A restock shows up as a button that stops being grey. A new location gets an Instagram post with the details inside an image. An event is a PDF flyer. Each of these is a real update that a person can see and a system has to infer. If the agent is comparing states over time, the source that states the change plainly, in text, with a date, is the one that is cheapest to quote.

This is an inference from how the feature is described, not a documented ranking rule. It does line up with what Google already documents for the systems these agents draw on.

#The signals Google already reads for change

Three of Google's existing pipelines are about detecting change, and each has documentation.

Product price and stock. Merchant Center automations update price, sale price, availability and condition by reading the structured data on your product pages, per Google's help page. The same page warns that automations "may not work for products with frequent price or availability value changes (for example more than once per day)" and are not a replacement for regular feed updates. For a "tell me when it drops below X" monitor fed by the Shopping Graph, a stale feed means your price change happened and the alert went to a competitor whose data was current.

Page updates. Google ignores changefreq and priority in sitemaps but uses lastmod "if it's consistently and verifiably" accurate, and defines a significant update as a change to main content, structured data or links, not a copyright year (Google Search Central). A sitemap that stamps every URL with today's date on every deploy teaches Google to discount the field on exactly the pages where you need it believed.

Openings. Google Business Profile lets a business set a future opening date up to a year out, and the profile appears on Google 90 days before that date (Google Business Profile Help). "New restaurants opening near me" is one of Google's own monitor examples. A pop-up with no profile and no opening date is relying on someone else's blog post to announce it.

#What to change

List the changes your customers would want to be told about. Restocks, price drops, launches, openings, schedule changes, new inventory, policy changes. For each, write down where that change currently appears on the web and whether it appears as text.

Publish each change as a dated, crawlable statement. A changelog, a "what's new" page, a press page or a product page section that says what changed and when. Keep the specifics (price, date, location, model) in HTML, not only in images or social posts.

Keep structured data and feeds in step with the page. Product markup with current price and availability, a feed that updates when inventory moves, and Merchant Center automations switched on as a backstop rather than the main mechanism.

Make lastmod honest. Update it only when content, structured data or links change. This is a small fix with a long tail.

Use the Business Profile fields that describe future states. Opening dates, special hours, temporary closures. These are the facts a local monitor is waiting for.

Measure what you can. You cannot see alerts. You can run your own monitors in AI Mode for the change queries that matter to you, record which sources each update cites, and log the lag between your change going live and the alert arriving. Repeat it across a few changes before drawing conclusions, since one alert is an anecdote.

#The open question

Google has not said whether monitors use the same retrieval as live AI Mode answers or a separate freshness-weighted path. Until it does, treat alert citations as their own series, separate from AI Mode answer citations. The pages that win one may not win the other, and the only way to find out is to watch both.