I turned off my Gemini history. What did I actually turn off?

In Transparency or stolen consent? I said the next part would be about Gemini, and what turning off its history really turns off. This is that part.

I turned it off

I turned off my Gemini history myself, because I didn't want my chats to be kept. Since then Gemini keeps none of my chats for me. If I close the tab, I can't get back to that conversation. When I open the link to an older chat, I just get an empty page telling me the history is off.

So far, that's exactly what I asked for. My question is what else sits behind that switch.

One switch, two jobs

The switch is now called Keep Activity.1 Here is its card:

The Keep activity card in Gemini settings, set to Off

Screenshot from October 29, 2026, from a personal account.

Read the first sentence. Keeping your activity "lets you pick up chats where you left off anytime and helps improve Google services, including AI models." Two jobs with two different beneficiaries, in one sentence and behind one switch. One is for me: remembering my chats. The other is for Google: learning from them.

Google's help page says the same thing. With the switch off, your future chats don't show up in your activity and aren't used to improve its models, unless you send feedback. (Gemini help) And the list of recent chats is only there when the switch is on. (recent chats help)

Now take the opposite case. Someone wants their history, so they can come back to old chats, but doesn't want those chats used for training. In a personal Gemini account we couldn't find any place that separates the two. Both, or neither.

The interesting part is that Google knows how to separate them. For Workspace2 accounts with enterprise protections, Google says chats aren't reviewed by people or used to improve its models, and conversation history is kept for 18 months by default. (Gemini help for work accounts)

ChatGPT separates them too. OpenAI's help page says it plainly: you can turn off "Improve the model for everyone" and keep your chat history. (OpenAI help) Funny enough, ChatGPT was the example of ambiguity in the previous post. Here it does better.

Does off mean nothing is kept?

No. The card itself says that when it's off, Google still saves chats for 72 hours to respond to you and help keep Gemini safe.

There's also something that isn't on the card, and you have to find it in the help page. Google says that even with the switch off, it may use human reviewers to help respond and keep things safe. And chats a person has reviewed are kept for up to three years, even if you delete them from your activity. Google does say it disconnects chats from your account before sending them to its service providers. (retention section of the Gemini help page)

Shahrzad’s note

Shahrzad:

To be fair, Google wrote the 72 hours right on that card and didn't hide it. It also gives the reasons in the help page: responding, safety, and not losing things if the system fails. And it added another path. Since August 2025 Gemini has Temporary Chat, a conversation that doesn't go into your history and isn't used to train its models, even when the switch is on. A few weeks later the switch was renamed from Gemini Apps Activity to Keep Activity. (Google's announcement)

But in what we read, we found no reason why history and training have to be one switch. And Temporary Chat doesn't help the person who wants a history. They should be able to say "keep it, but don't learn from it", the way you can in ChatGPT and in Google's own work accounts.

Being transparent isn't hard here either. One switch per choice. "Keep my history for me" is one. "Learn from my chats" is another. And next to each switch, in plain words, what still stays when it's off, who can read it, and for how long.

Notes (2)
  1. Formerly called Gemini Apps Activity. ↩︎
  2. Google Workspace, the business and school version of Google's services, like a company's Gmail and Drive. ↩︎