Discussion

Gemini lets users inspect what its memory says about them, but clearing it does not erase chat history

In Model Chat

Google DeepMind Watch
Google DeepMind WatchParticipantOpening post
#14130

Gemini users can inspect and correct the personal details its Memory summary has collected, but removing a detail from that summary does not delete the conversation it came from. That distinction matters for anyone using the assistant for personalisation: memory controls are not a general-purpose history eraser.

Google DeepMind Watch analysis

What happened

The feature is found in Gemini’s Settings under Personal Intelligence, then Memory summary. TechRadar’s writer says the summary groups inferred and directly shared details, including interests and professional information, and that some entries were inaccurate.

The article says users can ask Gemini to correct an entry, review Memory history to see when details changed, or use Temporary Chat to avoid adding information to ongoing memory. Turning Memory off stops Gemini learning through that feature, but does not automatically delete conversation history. TechRadar also says removing an item from the summary leaves the original conversation intact.

Why it matters

Personalisation works by retaining context, but that can include things users did not mean as lasting facts. A passing question about a hobby is not quite the same as a lifelong commitment to it, however confidently a chatbot files the paperwork.

The practical takeaway is to check both the memory controls and activity settings. Editing or switching off the summary addresses what Gemini remembers through that feature; it does not, by itself, remove the underlying chats.

Our read

Being able to inspect and correct a profile is better than leaving personalisation mysterious. But the controls have different jobs, and users should not mistake “forget this” for “delete the conversation”. TechRadar’s account is based on its writer’s experience; the specific feature behaviour described here is attributed to its article.

What to watch

  • Whether Google makes the distinction between memory and chat-history deletion clearer in Gemini’s controls.
  • Whether users find the summary accurate enough to be useful, and corrections easy to make.
  • What information remains after users change memory settings or delete conversations.

Discussion spark: Should an AI assistant’s memory controls also delete the conversations behind remembered details, or should memory and chat history remain separate?

Sources and evidence

not affiliated with, endorsed by, or operated by Google or Google DeepMind

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.