Google Information Agents: How AI Mode Keeps Watch for You
Google has opened AI Mode's "keep me updated" monitoring to everyone. Here is what these always-on search agents do, how to use them well and what they still can't promise.
In this article
- What are Google information agents?
- Who can use AI Mode monitoring now?
- How to set up an information agent in Google Search
- How do you manage or stop an alert?
- What can Google information agents track well?
- Information agents vs Google Alerts vs scheduled tasks
- What are the limits of Google information agents?
- Privacy: what you are telling Google
- Who should use it, and who can skip it
- Google information agents: FAQs
- Sources
Google information agents: the short answer
Google information agents are background tasks in AI Mode that keep searching after you close the chat. You ask AI Mode to “keep me updated on” something, and Google alerts you when it finds new, relevant information. They began with paid Google AI plans in summer 2026. On September 28, 2026, Google said the monitoring feature was rolling out to all users.
Google information agents turn a one-off search into a standing request. Instead of checking the same query every morning, you describe what you are waiting for, and Search keeps looking on your behalf. When something changes, you get an update with a short summary and links.
The idea is simple, but it matters. For most of the web’s history, “search” meant a question and an answer at one moment in time. Now Google is asking Search to remember what you care about and come back to you. That is a small but real step towards the always-on AI agents that every big tech company is chasing.
This guide explains what Google has announced, how to set up an alert, which tasks suit it and where the limits are. It is based on Google’s own announcements and reporting from the past few months; TheJusGrow has not tested the feature for this article.
What are Google information agents?
Google first showed information agents at its I/O developer conference on May 19, 2026. In its official I/O summary, the company said the agents operate “in the background, 24/7” and reason across information to find what you need.
In practice, an information agent is a saved instruction inside AI Mode, the chat-style version of Google Search. Google says it watches blogs, news sites and social posts. It also draws on Google’s own live data for finance, shopping and sports. When it spots a match, it sends you a synthesised update rather than a raw list of links.
Google’s two headline examples were apartment hunting and sneaker releases. For the first, you “brain dump” your exact requirements, such as budget, area and pets allowed. Then the agent flags listings that fit. For the second, you ask to hear when favourite athletes announce new shoe collaborations.
Who can use AI Mode monitoring now?
Access has widened in three stages. First, around June 12, 2026, 9to5Google reported that information agents were rolling out to Google AI Ultra subscribers. Google said the feature would reach Google AI Pro subscribers later in the summer. At that stage, it covered all AI Mode languages and markets for eligible subscribers.
Then, on September 28, 2026, Robby Stein, Google’s vice president of product for Search, said the monitoring feature was rolling out globally to all users. Search Engine Roundtable reported that it arrives through the Google app and may not appear for everyone at once. So if you don’t see it yet, that is expected during a staged rollout.
Google’s wording has also shifted. The spring announcements talked about “information agents”. The September update, however, mostly describes “info monitoring”. Both names refer to the same basic behaviour: AI Mode keeps checking a topic and tells you when something new turns up.
| Stage | Date (2026) | Who could use it |
|---|---|---|
| Announced at Google I/O | May 19 | Nobody yet; promised for summer |
| First rollout | Mid-June | Google AI Ultra subscribers |
| Paid expansion | Summer | Google AI Pro subscribers, per Google’s plan |
| Wide rollout | From September 28 | All users, gradually, via the Google app |
How to set up an information agent in Google Search
You don’t need a settings menu to start. Instead, you write a normal AI Mode request and add a monitoring phrase. According to 9to5Google, phrases such as “keep me updated on” or “alert me when” trigger an agent. Search Engine Journal adds that AI Mode replies by confirming it has set up a task to keep you updated.
- Open AI Mode in the Google app or on google.com, and sign in to your Google Account.
- Describe what you want to follow, with as much detail as you would give a helpful friend.
- Add “keep me updated” or “alert me when”, plus the condition that matters to you.
- Check the confirmation message, because it shows how AI Mode understood your request.
- Allow notifications from the Google app, since TechCrunch reported that updates arrive as push notifications.
Here are three example prompts you can adapt. Notice that each one states a clear condition, rather than a vague topic.
Keep me updated on two-bedroom apartments to rent in Austin under $2,200 a month that allow cats and are within a 20-minute walk of a train stop.
Alert me when tickets go on sale for family-friendly holiday light events within 30 minutes of Leeds, and tell me the price range.
Keep me updated if this blender comes back in stock or drops below $90 at a major retailer.
How do you manage or stop an alert?
When Google first described the feature, TechCrunch reported that active tracked topics appear in your AI Mode history. From there, you can jump back in to manage, refine or turn off an alert. That design matters, because a good alert often needs tuning after the first few updates.
Still, Google has not yet published a detailed help page for the global rollout. Search Engine Journal pointed out that the September announcement did not explain how often updates surface or how users switch monitoring off. So, for now, check your AI Mode history after setting an alert, and note where it lives.
If an agent sends too many updates, tighten the wording rather than deleting it at once. For example, change “new restaurants near me” to “new vegetarian restaurants opening within two miles of my postcode”. A narrower condition usually means fewer, more useful alerts.
What can Google information agents track well?
The best tasks share three traits. They involve information that changes, the change is public and the condition is easy to state. Google’s own examples fit that pattern, and so do several everyday uses.
| Good fit | Why it works | Tip |
|---|---|---|
| Back-in-stock and price drops | Draws on Google’s Shopping Graph, which Google says covers more than 60 billion product listings | Name the exact model and your target price |
| Rental listings | Listings change daily and criteria are concrete | Give budget, area, size and deal-breakers |
| Event and ticket announcements | Announcements are public and time-sensitive | Name the venue, date range and group size |
| New local openings | Google suggested restaurants and pop-ups nearby | Add cuisine, distance and opening hours |
| Sports and finance news | Google uses live sports and finance data | Say which team or company, and what counts as news |
For flights, Google also offers a dedicated option. Our guide to AI flight price tracking explains how AI Mode emails you when fares change. Meanwhile, shoppers who want to act on a price drop may find Google Universal Cart useful, because it gathers items from several retailers in one place.
Information agents vs Google Alerts vs scheduled tasks
Google Alerts has emailed people about new search results for over two decades. According to Google’s help page, you type a topic, choose options such as frequency and region, and receive emails when new results appear. It is reliable and simple, but it matches words rather than meaning.
By contrast, Google information agents try to understand a condition, such as “under $2,200 and allows cats”, and then summarise what changed. Scheduled tasks in ChatGPT, Gemini and other assistants work differently again. They run at set times, whether or not anything new has happened. We cover those in how scheduled AI tasks work.
| Google Alerts | Google information agents | Scheduled AI tasks | |
|---|---|---|---|
| How it starts | Keyword or phrase | Plain-English request in AI Mode | Prompt plus a schedule |
| When it runs | When new results appear | Ongoing monitoring, per Google | At fixed times |
| What you get | Email with links | Summary with links, via app notification | A written result each run |
| Best for | Mentions of a name or term | Changing conditions, such as stock or listings | Regular briefings |
For a wider look at Google’s personal agent plans, see our explainer on Gemini Spark. In short, AI Mode monitoring lives in Search, while Spark is pitched as a broader assistant.
What are the limits of Google information agents?
First, an alert is only as good as its sources. Google says the agents read blogs, forums and social posts as well as news sites. That breadth helps with early signals. However, it also means an update could rest on a rumour or a stale listing. Always open the linked source before you act.
Second, several practical details are still unclear. Google has not said how often agents check, how many alerts you can run or how it picks sources. Search Engine Journal flagged the same gaps. Therefore, don’t rely on AI Mode for anything where a missed update would cost you money, such as a visa deadline or a bill.
Third, an alert is not an action. Google says updates come “with the ability to take action”, such as following a link. Even so, the agent does not rent the flat or buy the blender for you. That keeps the risk low, but it also means you still have to move quickly.
Finally, keep the usual AI caution. Summaries can misread a page, and pages that talk to AI systems can try to steer them. Our guide to prompt injection explains why that matters for any agent that reads the open web.
Privacy: what you are telling Google
An alert is a standing record of what you want. For example, a rental alert reveals your budget and neighbourhood, and a ticket alert reveals your plans. Google’s announcements do not describe special privacy handling for these tasks. So it is reasonable to assume they sit with the rest of your AI Mode activity in your Google Account.
In practice, that means a few habits help. Avoid putting health details, full addresses or account numbers into an alert. Use a postcode area or neighbourhood instead of a street address. And review your saved alerts every few weeks, deleting the ones you no longer need.
Who should use it, and who can skip it
Google information agents suit people who are waiting for something specific and public. Renters, bargain hunters, parents watching for school-holiday events and sports fans are obvious examples. They also suit anyone who currently refreshes the same search every day.
On the other hand, you can skip them if Google Alerts already does the job, for instance when you only follow mentions of your name or business. Likewise, if you need guaranteed timing, a retailer’s own stock alert or an airline’s fare alert is the safer tool.
A quick checklist before you create an alert
- Is the thing you want public and likely to change?
- Can you state the condition in one or two sentences?
- Have you named places, prices and dates rather than general topics?
- Have you left out personal details the alert does not need?
- Do you know where to find the alert later to edit or remove it?
Key takeaways
- Google information agents keep watching the web after you leave AI Mode, then send a summary when something relevant changes.
- They launched for Google AI Ultra in June 2026, and Google said on September 28, 2026 that monitoring was rolling out to all users.
- Start one with phrases such as “keep me updated on” or “alert me when”, and state a clear condition.
- Google has not detailed check frequency, alert limits or source selection, so verify every update before acting.
- Keep alerts free of sensitive personal details, and prune old ones regularly.
Google information agents: FAQs
At first they required a paid Google AI Ultra or Pro plan. On September 28, 2026, Google’s Robby Stein said AI Mode monitoring was rolling out to all users, although it may not appear for everyone straight away.
Open AI Mode and add a phrase such as “keep me updated on” or “alert me when” to a detailed request. AI Mode should confirm that it has set up a task to keep you updated.
TechCrunch reported that the Google app sends push notifications, and that tracked topics appear in your AI Mode history, where you can refine or turn them off.
No. Based on Google’s descriptions, information agents send updates with links and follow-up options. They do not complete purchases or bookings on their own.
Sources
- Google, “Google Search’s I/O 2026 updates: AI agents and more” (May 19, 2026)
- TechCrunch, “How to use Google’s new AI agents to go beyond your standard searches” (May 19, 2026)
- Semrush, “Google Search expands agentic capabilities with information agents and Universal Cart” (May 21, 2026)
- 9to5Google, “Google AI Mode starts rolling out Search agents that keep track of information for you” (June 12, 2026)
- Search Engine Roundtable, “Google Search Rolls Out Information Agents In AI Mode For Google AI Ultra Subscribers” (June 15, 2026)
- Search Engine Roundtable, “Google Rolls Out Monitoring Capabilities In AI Mode To All” (September 28, 2026)
- Search Engine Journal, “Google Rolls Out AI Mode Info Monitoring To All Users Globally” (September 2026)
- Google Search Help, “Create an alert” (accessed September 29, 2026)



