On Gemini, a "temporary" chat now sticks around for two months
And the button that commits you the most isn't even in the settings.

The word that didn't change
On Gemini, next to "New chat," there's a "Temporary chat" button. You open it, ask your question, close it. Google's help page is plain about it: once you leave, you can't get back to it, and it never shows up in your history.
From where you sit, the chat evaporated. From Google's side, it stays on its servers for around two months. Until mid-November, that was 72 hours. The word "temporary" didn't change. The duration got roughly twenty times longer.
It's a bit like a coat check where you hand in your ticket on the way out, never knowing the coat stays behind the counter for two months.
Good news, with a line underneath
Google's announcement on November 16 is framed as a win for user control. Until now, one activity setting bundled two things: keeping your chat history, and letting Google use it to train its models. If you wanted one, you had to accept the other.
Now there are two switches. You can keep your chats without feeding the training pipeline. That's real, and people had asked for it.
The same post adds one sentence: if you turn the new History setting off, future chats will be saved for "around two months." It doesn't say it used to be 72 hours. You only find that out by comparing Gemini's privacy page with its archived copy from November 15, which is what TechCrunch and The Register did.
And the live page goes further than the blog post. It says: "Temporary chats and chats you have when your History setting is off are retained with your account for around two months." So the "temporary" button is covered, even though the announcement never mentions it.
"Off" depends on which switch
Google offers two justifications, and they aren't the same. The blog post talks about safety, as AI systems get smarter. The privacy page gives a more down-to-earth example: Google keeps chats for around two months "so they are available in case of a potential system failure." That line was already there at 72 hours.
What "off" means depends on the button:
- History off: chats stop appearing in your list, but they're kept for around two months and used to respond to you, process your feedback and "protect Google, its users, and the public," with help from human reviewers.
- Training off: your chats no longer train generative AI models. But the page says future chats "will be reviewed and used to improve Google services in other ways."
- Temporary chat: no training, but around two months of retention, the same as History off.
Opting out of training means opting out of one specific use, and only that one. Human review and the other improvements carry on. It's a switch that turns off one bulb in a room with several.
The quietest button
The heaviest lever isn't in the settings. It sits under every answer: the two little thumbs, "Good response" and "Bad response."
Google's help pages describe this as sending feedback. You click, and you can add a reason if you like. The related conversation goes along with it, plus files, images and content from connected apps. If History is off, Google also attaches context, including the last 24 hours of your chats.
That feedback is reviewed by human reviewers and kept for up to three years, separated from your Google account. Deleting your history doesn't change that: a chat that's already been reviewed stays. None of this is new, and the archived version already said three years. What changes is the contrast. You now have two switches to protect your conversations, and a tenth-of-a-second gesture that overrides both.
We found no measure of how many users know what this thumb triggers. It does look a lot like a "like" button on a social network, the kind of tap you make without thinking.
What the others do
An honest comparison doesn't put Google alone in the dock.
On temporary chats, the gap is clear. OpenAI says ChatGPT may keep a copy of a temporary chat for up to 30 days, for safety reasons. Anthropic keeps Claude's incognito chats for 30 days by default. Google, the shortest of the three at 72 hours, becomes the longest at around two months.
On the thumb, though, Google isn't the worst. Anthropic keeps data tied to a thumbs-up or thumbs-down for five years. OpenAI says that if you click a thumb after opting out of training, the entire conversation can still be used to train its models. The thumb is a side door across the whole industry. Google is simply the one that lengthened a duration this week.
And GDPR?
The European regulation sets out two principles that speak directly to this case: personal data must be "limited to what is necessary in relation to the purposes for which they are processed," and kept "for no longer than is necessary for the purposes for which the personal data are processed." One of the stated purposes is surviving an outage. A backup that lasted three days until mid-November now takes around sixty. Google did not answer TechCrunch and The Register when they asked what prompted the change.
We found no position from any authority on this change. Ireland's Data Protection Commission, Google's lead regulator in Europe, opened an inquiry in 2024 into an older Google model, PaLM 2; none of its press releases since June 2025 announces a decision. As for France's CNIL, we found nothing on Gemini.
On timing, Google says it will start rolling out the changes "globally" "in the coming weeks," with no country-by-country detail. Its privacy page, which has a chapter on European law, already shows the two months in its French version, with no regional exception.
What you can do about it
No setting brings the duration back to 72 hours. But a few habits change what leaves your hands and for how long.
Treat a temporary chat like a sent email, not a hallway conversation: anything you wouldn't write to a stranger doesn't belong there for two months. Save the thumb for ordinary conversations, since it attaches context and sends the chat to review for up to three years. And don't count on deletion to wipe what's already been reviewed.
In Gemini's interface, the button that commits you the most is also the smallest: a thumb under every answer.
Was this article useful?
Topics covered:
Frequently asked questions
How long does Google keep a Gemini temporary chat?
Why is Google keeping these chats longer?
What happens when you click the thumb under a Gemini response?
Does opting out of training protect your Gemini chats?
Does Gemini keep chats longer than ChatGPT or Claude?

Katja Liersch
Co-founder & Journalist
Katja is a journalist and TV producer. With decades of experience in mainstream media, she brings to Declic Media the perspective of those discovering AI: curious, demanding and pragmatic. She ensures every piece of content truly speaks to everyone.
All articles by Katja →