How to See What ChatGPT Knows About You
Open Settings, then Personalization, then Manage memories to read every fact ChatGPT has saved about you. You can also ask the model directly with a prompt like what do you know about me, review the separate Reference chat history setting, and pull a full archive through Settings, then Data Controls, then Export. Together these four checks show you the complete picture.
Two systems decide what ChatGPT knows about you
ChatGPT does not keep one single file on you. It uses two separate systems, and each is controlled in a different place. Knowing the difference is the first step to seeing the full picture.
The first system is Saved Memories: specific facts you or the model decided to keep, such as your name, your job, or a dietary preference. The second is Reference chat history: a broader signal where ChatGPT draws on patterns from your past conversations to make new ones more relevant. Saved Memories stay until you delete them. Details pulled from chat history can shift over time as the model updates what is useful.
- Saved Memories: discrete facts, listed and editable one by one.
- Reference chat history: looser context learned from past chats.
- Each has its own on or off toggle in Settings.
- Checking only one leaves part of the picture hidden.
View your saved memories in Settings
The fastest way to read the explicit list is through the menu. This shows every entry ChatGPT has filed away as a saved memory, with a control to remove any line you do not want kept.
On the web and desktop apps the path is the same. Mobile apps follow the same Settings then Personalization order with minor layout differences.
- Click your profile icon, then select Settings.
- Open Personalization in the sidebar.
- Select Manage memories to see the full list.
- Use the trash icon next to any entry to delete it.
- Use Clear ChatGPT's memory to wipe every saved item at once.
- Toggle Saved Memories off if you want the feature paused.
Ask ChatGPT directly what it knows
The settings list shows saved memories, but a direct prompt is a useful cross check and can surface how the model is interpreting what it holds. Type a plain question into any chat and read the answer against the Manage memories list.
Treat the reply as a summary, not a database export. The model paraphrases, and it may pull from active chat history as well as saved memories, so the wording will not always match the stored entries word for word.
- Ask: what do you know about me?
- Ask: list everything you have saved in memory about me.
- Ask: what preferences are you applying to my answers?
- Compare the reply to your Manage memories list for gaps.
Check Reference chat history separately
Reference chat history is the quieter influence. When it is on, ChatGPT can recall useful details from earlier conversations even when nothing was saved as a formal memory. If you once mentioned you like Thai food, it might factor that in the next time you ask what to cook.
Because this system does not produce a tidy list, the control that matters is the toggle. You manage it in the same Personalization area, and turning it off stops new answers from drawing on prior chats.
- Open Settings, then Personalization.
- Find the Reference chat history toggle.
- Turn it off to stop ChatGPT using past chats for context.
- Remember this is distinct from the Saved Memories toggle.
Export your full data archive
For the complete record, request a data export. This returns far more than the memory list: it bundles your chat history and other account data into a downloadable ZIP file, which is the most thorough way to audit what is tied to your account.
The request is processed by email. OpenAI notes the export can take time to arrive, and the download link expires within a day of delivery, so act on the email promptly.
- Open your profile menu, then select Settings.
- Select Data Controls.
- Under Export data, select Export.
- Select Confirm export on the confirmation screen.
- Watch for the email, then download before the link expires.
A memory layer built to be inspected: MemX
Reviewing ChatGPT's memory is a manual habit you have to repeat. MemX takes a different starting point: it is an external memory layer for your AI tools where seeing, editing, and exporting everything it holds is the default, not a settings hunt.
MemX is private by architecture, with per-user isolation, encryption at rest, and keys managed through Google Cloud KMS, plus on-device handling where it applies. It is a complement to ChatGPT, not a replacement. You keep using ChatGPT for the conversation while MemX gives you one transparent, portable record of what your assistant remembers.
- Every stored item is visible and editable by design.
- Export your full memory whenever you want, in a portable form.
- Private by architecture: per-user isolation and encryption at rest.
- Works alongside ChatGPT rather than replacing it.
Key takeaways
- ChatGPT splits what it knows across Saved Memories and Reference chat history, controlled in different places.
- Read your saved memories at Settings, then Personalization, then Manage memories, where you can delete entries.
- Ask the model what do you know about me as a quick cross check, but treat it as a summary not a full export.
- Reference chat history has no list, only a toggle in Personalization, so manage it on or off.
- Export the complete archive through Settings, then Data Controls, then Export, and download before the link expires.
Frequently asked questions
Sources
Skip the manual steps
MemX is an AI memory app: store anything, skip the folders, and find it again by asking in plain English. Private by architecture, with per-user isolation and encryption at rest.
Try MemX Free