AI Tools

Delete AI Memory: ChatGPT, Claude, Gemini

Arpit TripathiArpit TripathiLinkedIn·June 5, 2026·9 min read

See, edit, and delete what ChatGPT, Claude, and Gemini remember, plus the deletion myth nobody mentions.

You flip the memory toggle off, assume the slate is clean, and move on. It is not. Turning memory off does not erase what the tool already saved, and deleting a single memory entry does not touch what the model learned during training. To clear your saved memories you open each tool's memory list and remove the entries yourself, and even then a copy can sit in a safety window for days.

Below are the current menu paths for ChatGPT, Claude, and Gemini. Each section leads with the answer, then walks the clicks. At the end, three myths explain why "I deleted it" rarely means what people think it means.

Insight

Here is what most delete guides skip: a memory toggle controls whether the AI keeps learning. It does not wipe the profile it already built. Off is not delete. Hidden is not gone. Disabling and deleting are separate buttons, and you usually need both.

ChatGPT: Settings, Personalization, Manage memories

To delete a ChatGPT memory, go to Settings, then Personalization, then Manage memories, and remove entries one by one or clear them all. You can also tell ChatGPT "forget that I mentioned X" inside any chat, and it drops that detail and confirms.

ChatGPT splits memory in two. Saved memories are facts you explicitly asked it to remember. Reference chat history is the broader set of insights it pulls from past conversations on its own. Both live under Settings, Personalization, and each has its own toggle, per OpenAI's Memory FAQ.

Delete one memory or all of them

  • Delete a single entry: open Settings, Personalization, Manage memories, then delete any individual entry.
  • On mobile: tap and hold the entry, then select Forget.
  • Delete everything: in the Manage memories panel, use the clear-all option. This cannot be undone.
  • Conversationally: type "Forget that I mentioned [detail]" in any chat and ChatGPT removes it.
  • Check freshness: each saved memory shows the date it was added, so you can spot and remove stale entries.
Pro Tip

To fully remove something sensitive, delete the saved memory AND the original chat where you shared it. ChatGPT can rebuild context from old conversations if reference chat history is on, so clearing the memory list alone leaves a way back in.

Insight

Turning saved memory off does not delete anything already remembered. OpenAI states it directly: disabling memory stops new memories, but existing ones stay until you delete them. Deleted memories can be retained up to 30 days for safety and debugging.

sources

Claude: the one memory you can actually read

To view or delete Claude's memory, go to Settings, then Capabilities, then click View and edit memory. Claude shows a plain-language summary of everything it remembers, with citations linking back to the chats each fact came from. To erase it, toggle memory off and choose Reset memory.

Claude's approach is the most auditable of the three. You read the actual summary text instead of guessing at a hidden profile. Edit it by clicking the pencil icon in the lower left corner of the summary, or tell Claude in a chat what to remember or drop, per Anthropic's support documentation.

Pause versus reset

  • Pause memory: Claude keeps its existing memory but stops using it or adding new memories. Reversible.
  • Reset memory: permanently deletes all memories, including project memories. This cannot be undone.
  • Edit in place: click the pencil icon in the summary, or message Claude to add or correct a fact.
  • Incognito chat: click the ghost icon in the upper right of a new chat. Claude will not save it to memory or chat history.

Claude also imports and exports memory. Its import pulls a memory profile directly from ChatGPT, Gemini, or Grok, and you can export your Claude memory for backup or migration, via Settings, Capabilities, and the Memory section. Anthropic labels memory import as experimental and still in active development, so treat your exported file as the real backup.

Pro Tip

Want a clean profile but not a total wipe? Edit the summary text directly instead of resetting. Reset is all-or-nothing and takes project memories with it.

sources

Gemini: delete activity by time range or item

To delete Gemini's memory, go to gemini.google.com, open Settings and help, then Activity, and delete by time range or per item. Your personalization data lives in two places: Personal context (the memory toggle and saved instructions) and Gemini Apps Activity (your conversation history). You may need to clear both.

Gemini ties memory to your Google Account, so retention runs on a timer. By default, activity older than 18 months auto-deletes. You can change that window to 3 months or 36 months, or turn auto-delete off entirely, according to Google's Gemini Apps Help.

Where to click

  • Open gemini.google.com, choose Settings and help, then Activity (or go straight to myactivity.google.com/product/gemini).
  • Delete everything: click Delete, then All time.
  • Delete a slice: choose Last hour, Last day, or a Custom range.
  • Delete one item: use the delete control next to an individual entry.
  • Set retention: in the auto-delete settings, pick 3, 18, or 36 months, or turn it off.
Insight

Two retention catches most guides miss. Even with Keep Activity off, Google saves your conversations for up to 72 hours to run the service and process feedback. And any chat pulled for human review, plus related data, is kept for up to three years and is not removed when you delete your activity. "Delete" does not reach those copies.

Deleting Gemini Apps Activity leaves data that other Google services saved through Gemini, like Workspace or YouTube. Those follow their own policies. Many Gemini integrations also stop working once activity is off, since the assistant can no longer pull your personal data.

Temporary and incognito chats, vendor by vendor

Use a temporary or incognito chat for a one-off conversation that leaves no memory trace. All three tools offer one, and each keeps a short safety copy. None are truly off-the-record, but they do stop personalization.

  • ChatGPT Temporary Chat: does not use existing memories or create new ones, and is not used to train models. OpenAI may keep a copy for up to 30 days for safety.
  • Claude incognito: click the ghost icon in a new chat. The chat is not saved to memory or chat history.
  • Gemini Temporary Chats: do not appear in recent chats or Apps Activity, are not used for personalization, and are not used to train Google's models.

Read the safety window carefully. ChatGPT temporary chats can still draw on prior information for limited safety and security purposes, so "temporary" means "not personalized," not "never recorded."

Side by side: where memory lives and how to kill it

Quick reference: ChatGPT memory lives under Settings, Personalization; Claude under Settings, Capabilities; Gemini under Activity. Each has a separate delete control.

ControlChatGPTClaudeGemini
View what it knowsSettings > Personalization > Manage memoriesSettings > Capabilities > View and edit memorySettings > Personal context, plus Apps Activity
Delete allManage memories > clear allToggle off > Reset memoryActivity > Delete > All time
Edit a single factPer-entry delete, or say 'forget that'Pencil icon or ask in chatDelete individual activity items
Off-the-record modeTemporary ChatIncognito (ghost icon)Temporary Chats
Retention catchOff keeps existing memories; 30-day safety copyReset also wipes project memories72 hrs even when off; human-reviewed chats up to 3 yrs

Three things 'delete' does not do

Myth 1: turning memory off deletes it

Off means "stop learning," not "forget." OpenAI says plainly that turning saved memory off will not delete anything already remembered. Claude's pause keeps the existing memory intact. To erase, use the delete or reset action. That is a different button.

Myth 2: deleting the chat deletes the memory

These are separate stores. A fact ChatGPT saved to memory survives even after you delete the conversation that produced it, and the reverse holds too. That is why removing something sensitive often takes two deletions: the memory entry and the source chat.

Myth 3: deleting a memory un-teaches the model

Deleting a memory entry removes it from your personalization profile. It does not pull anything out of the model's trained weights. Once data shapes a model during training, the influence spreads across billions of parameters, and no delete button scrubs it out. Removing that influence takes machine unlearning or a full retrain, an active research area, not a settings toggle.

Insight

For everyday chats, your messages are not automatically training the base model the moment you send them, and consumer tools let you opt out of training. But anything a model already learned during its original training is a separate matter from your editable memory profile.

A cleaner option: memory you actually own

AI memory scatters across three vendors, two toggles each, a 72-hour holding window, and a human-review copy that can sit for years. After all that, you start to want a single store you control.

MemX (memx.app) is a personal AI memory app, a second brain you fill on purpose. Capture photos, PDFs, documents, voice notes, and WhatsApp messages, then ask in plain English. MemX answers from your own saved material and cites the source memory. It reads text inside photos with OCR, so a receipt or a whiteboard photo becomes searchable, not just a filename. Say a reminder out loud and it writes to your calendar, recurring included.

Privacy runs the opposite way from the toggle maze: your data is encrypted, never used to train AI, and off-limits to MemX staff. There is a free tier with no credit card on Android, iOS via TestFlight, and WhatsApp. For the deeper questions on portability and control, see our pieces on who owns your AI memory and whether you can export it, plus the hands-on comparison of every AI memory app we tested.

Written by Arpit Tripathi, who works on AI memory at MemX.

Frequently Asked Questions
01How do I delete ChatGPT memory?

Open Settings, Personalization, Manage memories. Delete entries individually, or use the clear-all option to wipe everything. You can also tell ChatGPT "forget that I mentioned X" in any chat.

02Does turning off ChatGPT memory delete what it already knows?

No. OpenAI states that turning saved memory off stops new memories but does not delete existing ones. To erase them, open Manage memories and delete the entries or clear all. Disabling and deleting are separate actions.

03Can I see everything Claude remembers about me?

Yes. Go to Settings, Capabilities, View and edit memory. Claude shows a readable summary with citations to the chats each fact came from. Edit it with the pencil icon, or choose Reset memory to permanently delete it.

04How long does Gemini keep my data after I delete it?

Gemini auto-deletes activity older than 18 months by default, adjustable to 3 or 36 months or off. Even with Keep Activity off, Google holds conversations up to 72 hours, and chats pulled for human review can be kept up to three years.

05Does deleting a memory remove my data from the AI model?

No. Deleting a memory removes it from your personalization profile, not from the model's trained weights. Data learned during training is spread across billions of parameters and only removable through machine unlearning or a full retrain.

Read Next

Or try MemX to access 40+ AI models in one place — including Claude Sonnet 4.6 and GPT-5.4 — and get your questions answered today.

Was this article helpful?

Found this useful? Share it with someone who needs it.

Free · iOS, Android & WhatsApp

Stop losing what you save.
Let MemX remember it for you.

Every screenshot, photo, PDF and voice note — captured, encrypted, and instantly searchable. Ask in plain English, get the answer in seconds.

  • Reads text inside images and handwriting
  • Private and encrypted by default
  • Free to start, no credit card

Takes under a minute to set up. Your data stays yours.

Arpit Tripathi
Written by
Arpit TripathiLinkedIn

Founder of MemX. Ex-Google Staff Tech Lead Manager, ex-AWS Senior SDE (Elastic Block Store). Writes about practical AI on the MemX blog.

Keep reading

More guides for AI-powered students.