Google Wants You to Pay $100 a Month to Let It Watch You All Day

Google announced something at I/O that deserves more scrutiny than it has gotten: information agents, a new product that monitors the web on your behalf, around the clock, and sends you push notifications when something matches what you care about.

The announcement was easy to file under AI feature. But read it again. Google is selling a product that watches everything on the web, all the time, specifically on your behalf. It sent the press release on a Tuesday. Nobody asked what it does with that data.

Here is what Google says the product does. According to the official Google blog and confirmed by vice president of product Robby Stein in WIRED, information agents operate in the background, 24/7, monitoring blogs, news sites, social posts, and real-time data including finance, shopping, and sports. You define what you want tracked. The agent watches for it, synthesizes relevance, and pushes an alert when something matches. Stein example: you could be asleep, and the agent is still working for you. The product launches this summer for Google AI Pro and Ultra subscribers at $19.99/month and $99.99/month respectively. Free users get access, Google says, when the time is right. There is no defined timeline.

The pricing structure is worth getting right because Mycroft already filed a story correcting the same error: Google did not drop its AI Ultra price from $250 to $100. It introduced a new $100/month entry tier and reduced the existing top-tier from $250 to $200. Two separate actions. The floor for accessing information agents is $100/month.

Now here is the accountability question that has not been answered. Googles existing business runs on surveillance. Five billion users search for free; Google monetizes the attention through advertising. The data trail from search queries is the asset. Information agents represent a qualitatively different kind of surveillance. They are not responding to a query. They are running continuously in the background, monitoring specific topics you defined, across the entire web. That data — what you chose to track, how often, what triggered an alert, what did not — is a new kind of behavioral signal Google does not currently have from search.

The WIRED coverage noted these agents keep users trapped within Googles ecosystem without browsing the web themselves, similar to how AI Overviews worked when launched. But the ecosystem trapping is almost beside the point. The point is the monitoring surface area. A search query tells Google what you thought to type. An information agent tells Google what you cared enough about to set a persistent, automated watch on. Those are different signals with different value.

TechCrunch compared information agents to the next evolution of Google Alerts, the notification service Google launched in 2003. That comparison is instructive but incomplete. Google Alerts sends you an email when your name appears on a website. Information agents are supposed to synthesize across sources, reason about relevance, and take action on your behalf — booking a barber, tracking a sneaker drop, monitoring a competitors product page. The agent is not just watching. It is deciding what matters and what to do next.

Googles own blog confirms Gemini Spark, a companion product that reads your Gmail, surfaces trip plans, and tracks what needs restocking, is limited to Ultra subscribers at launch. The same tier that gets information agents. If you want Google reading your inbox to manage your life, the price is $100/month minimum.

Whether that trade is worth it depends on an answer nobody has provided: what happens to the data these agents gather? Google has not published a detailed data retention policy for information agents specifically. The general Google AI privacy terms apply, but the continuous, proactive nature of monitoring creates different data patterns than reactive search. Every topic you track, every alert that fires, every web page the agent visits to check your conditions — each is a data point.

This is the gap in the coverage. The pricing story was covered. The feature announcement was covered. The accountability question — what Google is now in the business of selling, what it does with that data, and whether the privacy terms actually constrain what the product can do — has not been.

A reporter with a Google AI Pro or Ultra subscription could answer part of that question by activating an information agent and observing, through browser dev tools and network traffic analysis, what the agent actually accesses, how often it calls Googles servers, and what data leaves the device. That is the reporting that would separate this from a feature announcement.

What can be said now, without that test: Google is offering a 24/7 surveillance subscription. The surveillance model that built a $2 trillion company is now being sold as a $100/month add-on. The question of what Google does with that new continuous behavioral signal — whether it trains models on it, whether it monetizes it separately from advertising, whether it retains it indefinitely — has not been answered in any public documentation.

Until someone answers it, the story is not complete.