After months of conversations, ChatGPT knows a lot about you. Your job, your people, your plans, the thing you were worried about in March. Most people have never looked at what is in there. Here is how to look, and what to do with what you find.
Step 1: ask the question
Open a new chat and type:
what do you know about me? write out everything, organized by topic.
The answer is usually surprising in both directions. It knows things you forgot telling it. It also has things wrong, sometimes importantly wrong, and it is missing things you would have bet it knew.
Two follow-ups get more out of it:
what have I told you about the people in my life?
what projects and decisions of mine do you know about?
Step 2: look at the official panel
Settings, then Personalization, then Memory. This shows the notes ChatGPT explicitly saved. Compare it with the answer from step 1 and you will notice they do not match. That is because memory lives in more than one layer: saved notes, summaries of chat history, and whatever the current model retrieves. There is no single place where you can see all of it, which is worth knowing before you rely on it.
Step 3: save the answer somewhere you control
Copy the full answer from step 1 into a note, a doc, anywhere outside ChatGPT. Do this today. In August 2026, a wave of users opened the memory panel and found it blank, and OpenAI's status page acknowledged the incident. The people who lost the least were the ones with a copy.
For a fuller backup: Settings, then Data Controls, then Export, which emails you your entire chat history.
Step 4: turn the snapshot into memory you own
Here is the thing about the copy you just made: it is a snapshot. It stops updating the moment you paste it into a note. What you actually want is for that knowledge to keep growing, somewhere you can always read it.
That is the product we build. Talkamore takes that paste as a starting point: give it what ChatGPT wrote about you, and it turns it into pages, one per person, project, and topic in your life. From then on you just talk to it, and the pages write themselves. Every page is yours to open, read, and correct, and every answer it gives shows the day you said the thing.
The difference from where you started: ChatGPT's memory is a locked box that answered one question today. Your Talkamore pages are a record you can open on any day, and losing them is not a thing that a bad Tuesday at a company can do to you. When a page fills up, older notes move to an archive page instead of disappearing, and that is enforced in code.
It also stops the knowledge being trapped in one app: connect Claude, Claude Code, or Cursor and they read and write the same pages.
The five-minute version
- Ask ChatGPT: "what do you know about me? write it all out."
- Copy the answer into somewhere you control.
- Paste it into talkamore.com when it asks what your AI already knows. Free to start.
- Keep talking. The memory keeps itself from here.
Whatever you do with step 3, do steps 1 and 2 today. You spent months teaching an AI who you are. The least it owes you is a copy.