Google’s AI has been quietly taking notes on you, and until this week you couldn’t do much about it. That just changed: Gemini memory controls finally let you see what the chatbot remembers, fix what’s wrong, and delete what you’d rather keep private. Here’s how to take that control in about two minutes.
What Gemini memory actually is
Gemini memory is the feature where the chatbot saves details about you between chats. Your job title, your projects, how you like your answers formatted, that trip you keep planning. Instead of re-explaining yourself every time, the assistant pulls from this running summary.
The catch: for months, Gemini users couldn’t inspect it. ChatGPT and Claude both let you open a memory summary and edit it. Gemini just remembered things and left you guessing. One MakeUseOf experiment made the gap obvious: the writers taught ChatGPT, Claude, and Gemini the same false fact, then tried to erase it. Gemini held onto the false info while the other two showed editable summaries.
Google caught up. The new Memory summary feature opens the notebook. And once you can read it, you can rewrite it.
How the three big chatbots compare
Here’s where each assistant stood as of this week, based on Google’s documentation and the MakeUseOf experiment:
| Assistant | View what it remembers | Edit or delete entries | How |
|---|---|---|---|
| ChatGPT | Yes | Yes | Memory summary in settings, editable |
| Claude | Yes | Yes | Editable memory summary |
| Gemini | Yes, new | Yes, new | Settings > Personal Intelligence > Memory summary |
That middle column is the headline. Gemini was the odd one out for months, and now the three behave alike. The new Gemini memory controls put Google’s assistant on the same footing, so you can run the same privacy audit on each.
Where to find the Gemini memory controls
Two routes, both take seconds.
Route one, ask directly. In the Gemini app, type the question below and the summary appears in chat:
What do you know about me?
Route two, dig into settings. Go to Settings, then Personal Intelligence, then Memory summary. Same information, one screen. Google’s own Memory summary support page confirms both paths.
From there, you manage everything through written prompts. Google’s examples are refreshingly blunt: “Add to my memory that I speak Japanese” adds a fact, and “Remove my old phone number from your memory” kills one. You can also correct a wrong entry (I moved, my city is Austin now) or order a bigger deletion (forget about my job search). Type the change, confirm it, done. Gemini may ask a follow-up question if your request is ambiguous.
Want the whole thing off? The main Personal Intelligence menu includes a switch that disables Memory entirely.
Six limits you should know upfront
Before you go clean house, a few catches. These are the details most guides skip:
- It’s US-only for now, and English-only. Elsewhere, the feature hasn’t landed yet.
- You need a paid Google AI plan. Free accounts don’t get Memory summary.
- Both toggles must be on: Keep Activity and Memory, in that order of importance.
- Work and school accounts are excluded. Google limits this to personal accounts, and you need to be 18 or older.
- Gems and Live chats don’t support it. The summary works in regular chats on the mobile and web apps.
- It skips other surfaces. If you use Gemini through Workspace tools or third-party apps, look elsewhere.
There’s a subtler point too. The summary you can edit is not a complete record of everything Google holds. Gemini’s memory, your chat history, Connected Apps, and the activity saved to your Google account are four related but separate systems. Editing the summary doesn’t touch the others.
The summary is not the whole file
That last point deserves its own section, because it’s where people get a false sense of privacy.
Think of Memory summary as the assistant’s cheat sheet, not the archive. The archive includes your past chats and, if you’ve granted access, data from Gmail, Drive, Photos, Search, and YouTube. Gemini’s personalization blends all of that to tailor responses. You can correct the cheat sheet in seconds. The archive has its own settings buried deeper in your Google account.
So if privacy is the goal, don’t stop at the Memory summary. Review your chat activity controls too. We walked through a similar audit for ChatGPT’s training setting, and the same habit applies here: check what’s being stored, decide what shouldn’t be, act once.
What else Google holds, and how to manage it
Since the summary is only the cheat sheet, here’s the map to the rest of the filing cabinet. All of it lives in your Google account, and all of it is worth a one-time cleanup:
- Chat activity: your My Activity page lists every Gemini conversation. You can delete individual chats or wipe ranges of history by date.
- Connected Apps: the apps you’ve granted access to Gemini keep working until you revoke them. Audit the list once, and cut anything you don’t recognize.
- Memory itself: the toggle in Personal Intelligence turns the whole feature off if editing feels like too much work.
- The broader personalization: responses can draw on Gmail, Drive, Photos, Search, and YouTube when you’ve enabled those connections. Each connection has its own switch in Google account settings.
Ten minutes here does more for your privacy than any VPN. And unlike the archive, the Gemini memory controls give you instant feedback: change the summary, and the next answer reflects it.
Why this matters more than it sounds
Memory isn’t only about privacy. It’s about answer quality.
A slightly weaker model that knows your context beats a stronger one starting from zero. That’s true for work (my audience is beginners, skip the jargon) and for life (we’re vegetarian, plan accordingly). When the memory is wrong, every answer downstream inherits the error. When it’s stale, you get suggestions for a job you left two years ago.
Auditing it is basic hygiene now. Meta users got the same power earlier this year: our Meta Muse privacy guide covers seeing and deleting what that system knows. Expect the rest of the industry to follow, because nobody wants to be the assistant that won’t forget.
One practical tip: ask the question once a month. The Gemini memory controls only work if you actually use them, and a two-minute check tells you exactly what your AI thinks it knows.
If you’d rather not depend on any company’s memory at all, there’s a third option. We built a DIY AI memory system for ChatGPT and Claude that keeps the facts in a file you own. Gemini users can borrow the same trick.
Your next move
Open Gemini right now and ask what it knows about you. If the summary is accurate, you’re set. If it’s not, correct it before your next real question, because every chat after that one inherits whatever’s in that file.