CORRECTING CHATGPT MEMORY

ChatGPT remembered something wrong about me. How do I fix it?

Open Settings, then Personalization, then Memory. Delete or correct the wrong saved memory, and type a correction into the memory summary or highlight the wrong text. You can also tell ChatGPT the detail is incorrect or outdated. Deleting the chat it came from does not necessarily remove a separate saved memory.

Last verified

OpenAI: Memory in ChatGPT. Verified .

OpenAI Help Centre: Memory in ChatGPT. Verified .

Where do I see what ChatGPT remembers about me?

Open Settings, then Personalization, then Memory. Available controls may include Memory, Reference saved memories, Reference chat history, Memory summary, Manage, or Saved memories, varying by plan and platform. You can ask ChatGPT what it remembers. The memory summary does not necessarily include every detail. Enter a correction in its text field, highlight the wrong text and provide a correction, or tell ChatGPT the detail is incorrect or outdated. You can edit the memory summary or a saved memory when those controls are available. Saved-memory controls can include version history depending on plan and platform. OpenAI: Memory in ChatGPT. Verified .

How do I correct a wrong memory, step by step?

Open Settings, then Personalization, then Memory. Available controls may include Memory, Reference saved memories, Reference chat history, Memory summary, Manage, or Saved memories, varying by plan and platform. You can ask ChatGPT what it remembers. The memory summary does not necessarily include every detail. Enter a correction in its text field, highlight the wrong text and provide a correction, or tell ChatGPT the detail is incorrect or outdated. You can edit the memory summary or a saved memory when those controls are available. Saved-memory controls can include version history depending on plan and platform. OpenAI: Memory in ChatGPT. Verified .

  1. Open Memory

    Open Settings, then Personalization, then Memory.

  2. Correct the memory summary

    Enter a correction in the text field or highlight the wrong text.

  3. Delete or edit the wrong saved memory

    Open saved memories and delete the wrong entry, or edit it where that control is available. OpenAI says you can edit the memory summary or a saved memory when those controls are available.

  4. Tell ChatGPT the detail is wrong

    In a chat, say that the detail is incorrect or outdated.

  5. Check your correction

    Ask ChatGPT what it remembers about that topic and check the answer matches your correction.

Why does the wrong detail keep coming back?

Deleting a chat alone does not necessarily delete a separate saved memory created from that chat. OpenAI Help Centre: Memory in ChatGPT. Verified .

The memory summary is a high-level view of information ChatGPT may use to personalise responses. You correct it by typing a correction or highlighting text. Saved memories are details you ask ChatGPT to remember, or that it saves as useful context, stored separately from chat history. OpenAI Help Centre: Memory in ChatGPT. Verified .

Don't mention this again reduces future references to the information. It does not delete the original source. OpenAI: Memory in ChatGPT. Verified .

Information derived from chat history can change as ChatGPT updates what is most useful to remember. OpenAI: Memory in ChatGPT. Verified .

OpenAI may keep logs of deleted saved memories for up to 30 days for safety and debugging. Turning off reference chat history schedules what it remembered from past chats for deletion within 30 days; the chats themselves stay in your history. OpenAI Help Centre: Memory FAQ. Verified .

User report, not OpenAI policy: in community thread 1393739, a Plus user reported on 31 August 2026 that saved instructions could be recalled but were not consistently applied. OpenAI Support replied on 18 October 2026 with suggestions it said "may improve consistency, but they are not a confirmed resolution". OpenAI community: recalled memory not consistently applied. Third-party source. Verified .

What happens when two ChatGPT memories contradict each other?

Delete the outdated entry rather than adding a second one beside it.

Which memory wins when two ChatGPT memories contradict each other: not published. OpenAI says ChatGPT decides which available information is relevant to a response. OpenAI: Memory in ChatGPT. Verified .

User report, not OpenAI policy: community thread 1144222, posted in March 2025, reports contradictory memory and context across chats. The poster says the bug report was written by GPT-4.5; its explanation is not a verified technical diagnosis. OpenAI community: memory and context contradiction (March 2025). Third-party source. Verified .

Can I see what a memory said before it changed?

OpenAI says saved-memory controls can include version history depending on plan and platform.

Open Settings, then Personalization, then Memory. Available controls may include Memory, Reference saved memories, Reference chat history, Memory summary, Manage, or Saved memories, varying by plan and platform. You can ask ChatGPT what it remembers. The memory summary does not necessarily include every detail. Enter a correction in its text field, highlight the wrong text and provide a correction, or tell ChatGPT the detail is incorrect or outdated. You can edit the memory summary or a saved memory when those controls are available. Saved-memory controls can include version history depending on plan and platform. OpenAI: Memory in ChatGPT. Verified .

How to audit ChatGPT memory and version history

How do I stop a new wrong memory being saved?

Choose Unpersonalized before starting if you want no memories applied.

Temporary chats do not create or update memories. Since 27 August 2026 you choose when you start one whether it uses your existing memories and custom instructions, and that cannot be changed after the chat begins. The 27 August 2026 change is recorded in the ChatGPT release notes (help.openai.com/en/articles/6825453-chatgpt-release-notes). OpenAI Help Centre: Memory FAQ. Verified .

Temporary chat memory and privacy

How does AI Twin handle a wrong memory?

Review before anything is kept: Nothing is saved until you confirm. Duplicates and conflicts come to you to decide.

Duplicates and conflicts: Your Twin spots the same person twice, and two facts that disagree. You choose: merge, keep both, or keep the newer one.

Typed Memory: Every memory entity carries a category: People, Documents, Facts, Events, Episodes, and further types beneath them. Queryable, editable, versioned, audited.

Universal Audit: Every meaningful action is logged in an append-only audit log you can read.

What changed

20 November 2026: some guides say ChatGPT saved memories can only be deleted, not edited. OpenAI's help page now says you can edit the memory summary or a saved memory when those controls are available.

Deleting a chat does not delete its memories: OpenAI says deleting a chat alone does not necessarily delete a separate saved memory.

Open Settings, then Personalization, then Memory. Available controls may include Memory, Reference saved memories, Reference chat history, Memory summary, Manage, or Saved memories, varying by plan and platform. You can ask ChatGPT what it remembers. The memory summary does not necessarily include every detail. Enter a correction in its text field, highlight the wrong text and provide a correction, or tell ChatGPT the detail is incorrect or outdated. You can edit the memory summary or a saved memory when those controls are available. Saved-memory controls can include version history depending on plan and platform. OpenAI: Memory in ChatGPT. Verified .

Deleting a chat alone does not necessarily delete a separate saved memory created from that chat. OpenAI Help Centre: Memory in ChatGPT. Verified .

Questions people ask about AI memory

Can I edit a ChatGPT saved memory or only delete it?

Open Settings, then Personalization, then Memory. Available controls may include Memory, Reference saved memories, Reference chat history, Memory summary, Manage, or Saved memories, varying by plan and platform. You can ask ChatGPT what it remembers. The memory summary does not necessarily include every detail. Enter a correction in its text field, highlight the wrong text and provide a correction, or tell ChatGPT the detail is incorrect or outdated. You can edit the memory summary or a saved memory when those controls are available. Saved-memory controls can include version history depending on plan and platform.

OpenAI: Memory in ChatGPT. Verified .

Does deleting the chat remove the wrong memory?

Deleting a chat alone does not necessarily delete a separate saved memory created from that chat.

OpenAI Help Centre: Memory in ChatGPT. Verified .

What does "Don't mention this again" do?

Don't mention this again reduces future references to the information. It does not delete the original source.

OpenAI: Memory in ChatGPT. Verified .

How long does OpenAI keep a deleted memory?

OpenAI may keep logs of deleted saved memories for up to 30 days for safety and debugging. Turning off reference chat history schedules what it remembered from past chats for deletion within 30 days; the chats themselves stay in your history.

OpenAI Help Centre: Memory FAQ. Verified .

Do temporary chats change my memory?

Temporary chats do not create or update memories. Since 27 August 2026 you choose when you start one whether it uses your existing memories and custom instructions, and that cannot be changed after the chat begins. The 27 August 2026 change is recorded in the ChatGPT release notes (help.openai.com/en/articles/6825453-chatgpt-release-notes).

OpenAI Help Centre: Memory FAQ. Verified .

What we checked

Checked . Read directly on 20 November 2026. The requested help-lb.openai.com mirror returned the help directory, not the article. The primary article showed Updated: 7 hours ago; no absolute update date is inferred.

OpenAI: Memory in ChatGPT. Verified .

OpenAI community: recalled memory not consistently applied. Third-party source. Verified .

OpenAI community: memory and context contradiction (March 2025). Third-party source. Verified .

AI memory

AI Twin is a personal AI memory layer, built in the UK. It keeps what matters to you as typed, searchable memory stored in the UK, and assistants that support MCP connectors, such as Claude, can read it with your permission.