Why Does ChatGPT Forget Everything Between Conversations?

Last updated: June 24, 2026

The short answer: it's the context window

When you type to an AI like ChatGPT, the model reads everything in the current conversation — your messages, its replies, any files you pasted — into a temporary working memory called the context window. It reads that window, generates a response, and that's it. When you close the conversation, the window is cleared.

The next time you open a fresh chat, the window is empty again. The model has no record of the last conversation. It isn't ignoring you and it isn't broken — it genuinely has nothing to read. Everything from before is gone unless you paste it back in.

Why this hits busy parents hardest

If you only have a 45-minute nap window to work, this is the tax that quietly eats your time. You open a new chat, and before you can do anything useful you have to re-explain: who you are, what you're building, what you decided last session, what you've already tried. Ten minutes gone before the real work starts.

Across five sessions a week, that re-onboarding adds up to most of a working hour spent telling the AI things you already told it. The frustration isn't that AI is unhelpful — it's that you keep paying the same setup cost over and over.

Doesn't ChatGPT have a memory feature now?

It does, and it helps a little. ChatGPT can save short personal facts and reuse them. But it's automatic and limited: it decides what to keep, it stores a small amount, and you don't control exactly what reloads into a given conversation. For a one-off preference, that's fine. For a multi-week project with decisions that evolve, it's too thin and too uncontrolled to lean on.

What you actually want is full control: a record of your project that you write, that comes back in full every session, that you can edit as things change.

The fix: a context file the AI reloads every time

The reliable way to give an AI memory across sessions is to put your context in a plain-text file and use a tool that reads that file automatically at the start of each session. Instead of the model starting from a blank slate, it opens already knowing your context.

One common version of this is a CLAUDE.md file — a configuration file that Claude Code reads at the start of every session. You write it once, keep it updated, and the AI loads it each time:

1. Write a plain-text file with who you are, what you're building, and your key decisions and preferences.
2. Use a tool like Claude Code that reads that file automatically when a session starts.
3. Add new decisions to the file as you go, so each session reloads the latest version — not a blank slate.

It doesn't change how the context window works. It just means the window gets filled with your context for you at the start, instead of you re-typing it every time.

The mechanism in one line

AI doesn't forget because it's careless — it forgets because each conversation is a closed box, and nothing carries over unless you carry it. Give it a file to read, and the forgetting stops being your problem.