How to Edit or Wipe ChatGPT's Automatic Memory — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- ChatGPT introduced memory in April 2024 requiring explicit "remember this" commands, but a June update added automatic "dreaming" that builds, edits, and sifts through memories without user prompts
- OpenAI says memory's three stated aims are carrying forward useful context, following user preferences and constraints, and staying current over time
- ChatGPT can form inaccurate assumptions from conversation patterns — such as labeling a user a DIY enthusiast after a week of light-switch research or believing they relocated after vacation-related searches
- Users can access memory controls via Settings > Personalization > Memory on the web app, or menu > profile > Memory on mobile, where they can highlight entries and choose "Make a correction" or "Don't mention this again"
- ChatGPT's memory toggle stops the AI from referencing stored data but does not delete it — full erasure requires selecting the three-dot menu and choosing "Delete and turn off memory"
- Temporary chats remain entirely outside memory — nothing is committed, and conversations are forgotten as soon as the chat is closed
Why it matters: With memory now forming automatically rather than only on explicit user request, users have less direct say over what ChatGPT retains — and those stored assumptions bleed into every future response. The toggle labeled "turn off memory" stops the AI from using stored data but leaves it on OpenAI's servers, meaning the only real reset is the full delete option in the three-dot menu.
Ask SkimNews
