Google has launched a new Memory summary feature for Gemini — giving users in the US a clear, editable list of what the AI remembers about them: background, relationships, work, interests, and preferences. It’s live now in Gemini’s desktop and mobile browser versions, but not yet in the standalone Gemini app, Tom’s Guide confirmed.
The feature is available to users aged 18 and over with an active Google AI plan, or, as Tom’s Guide observed while testing with a personal account, sometimes even on the free tier, provided both ‘Keep activity’ and ‘Memory’ are enabled. You’ll find it under Settings > Personal Intelligence > Memory > Memory summary.
From there, you can edit or delete any remembered detail, or turn off memory entirely. This isn’t just a dump of chat logs. It’s a distilled, human-readable profile Gemini built from your past conversations. No third-party apps, no hidden inference layers: just what you’ve told it, organized and surfaced.
And yes, you can add things it missed. Like your job title. Or your stance on AI in search. Or why you’d rather use DuckDuckGo than let a model remember your lunch order.
The rollout is global in scope but region-locked: only US users see it so far. No date has been given for expansion to India, the UK, or elsewhere. Google hasn’t said whether the Memory summary will eventually appear in the Android Gemini app, or how it’ll interact with Android’s upcoming developer verification rules, which now apply to all apps, not just Play Store ones.
That overlap matters. Because if Gemini’s memory ties into your Google Account, and your account ties into every Android app you install, then what you edit here may ripple further than chat history. For now, though, it’s simple: a single page. A few bullet points.
And the first time Google has let you read, revise, or erase its working model of you. The feature is rolling out now. According to Tom’s Guide’s Elton Jones, who covers AI for the outlet and tests models including ChatGPT, Gemini, and Claude, the Memory summary doesn’t show raw data, it presents a curated, point-by-point summary.
He tested it using his own Gemini account and found it listed details like his professional background and interests, drawn solely from prior chats. He noted he could add more context, like his specific profession, at any time. Jones also emphasized that this level of transparency, letting users see what the AI has learned, not just that it remembers, is rare among consumer-facing large language models.
Most competitors still offer only deletion options, not editing or review. And while Google says the Memory feature is tied to the Google AI plan, Jones found it accessible without one, suggesting the company may be soft-launching access ahead of formal eligibility rules. Still, the limitation remains firm: no Android app support yet, no non-US availability, and no public timeline for either.
What comes next? Either wider rollout, or deeper integration with Android’s broader identity and permissions architecture. The feature is rolling out now.
ADFiled by The AI Desk
Models, assistants and the companies and chips behind them, reported from what was released and what was claimed, with the difference kept clear.
More from this desk →
Be the first to comment
Join the argument. No password, just your email or a passkey.