Google has rolled out memory controls for Gemini in the US — giving paid subscribers a way to see, edit and delete what the AI knows about them. The feature is live as of Tuesday morning, November 18, 2026, and requires both a Google Account and an active Google AI subscription (the $19.99/month plan).
It is not available to free users or to anyone outside the US at launch. The rollout is web-only for now: no Android or iOS app support yet. No timeline has been given for expansion to other markets or platforms. Gemini’s memory isn’t a log of chats.
It’s a distilled profile, built only if Keep Activity is turned on, that pulls recurring facts from past conversations: your job title, your pet’s name, dietary restrictions, travel preferences, even your preferred coffee order. That profile lives separately from chat history and shapes how Gemini answers questions across sessions. You can view it in the Gemini web app under Settings > Personal Intelligence > Memory summary.
What appears is a bulleted list, grouped into categories like “Personal data”, “Career”, and “Interests”. Each line is editable, click the pen icon, rewrite or delete, then hit Save. You can also prompt Gemini directly via the text box at the bottom: “Forget I mentioned my old address” or “Add that I use Vim, not VS Code”.
Changes are logged. Gemini keeps a memory history, timestamps included, of every edit you make. Turning off the Memory toggle stops new facts from being added but doesn’t erase existing ones. To scrub those, you must either edit them individually or delete the source chats.
Those live in Settings > Activity, where you can scroll by date and remove any conversation with the “X”. Deleting a chat removes it from Gemini’s context pool, but won’t retroactively purge facts already baked into the memory summary unless you manually delete them there too. This isn’t privacy-by-default. Keep Activity is on by default for users 18 and older.
Memory is off by default, but only if you haven’t already enabled it. Once toggled on, it starts building without further prompts. The controls respond to growing scrutiny over AI memory practices, not just from regulators, but from users searching “google gemini ai” and “google gemini memory” at record volume this month.
They’re also a direct answer to complaints that Gemini’s personalization feels intrusive when uncorrectable. Now it’s correctable. Just not universally. Not yet.
The feature rolls out globally “in the coming weeks”, Google says, but no country or region beyond the US has been named. No price or eligibility details have been shared for non-US users. For now, if you’re not in the US and paying $19.99/month, you can’t see what Gemini remembers, or change it.
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.