ChatGPT
Updated Sep 28, 2026
ChatGPT is the AI assistant most people have used, and it keeps every conversation you have ever had with it, including the ones you forgot about. The export is a ZIP from Settings, and the file worth opening first is chat.html, which lays your whole history out in a browser with no tools and no internet connection. Nothing in the archive can be imported back into ChatGPT, so it is the only copy there will ever be.
What you'll get
Before you start
- Read Temporary Chat before you start. Those conversations were never saved to your history, so they are not in the export and there is no way to get them back. Anything you want from a temporary chat has to be in a normal conversation first.
- Have a look at what you have. A decade of ChatGPT is a lot of files, and it is worth knowing whether you want the lot or only the projects you still care about.
- Work out where the archive is going, because the download link in the email is good for 24 hours and then it is gone. Requesting a second one is easy, so this is only a problem if you were planning to do it months from now.
How to download
- Go to Settings, then Data controls, then Export data.
- Click Export and confirm. There is a toggle for whether to include conversations you have already deleted, and it is worth thinking about rather than leaving on whatever it defaults to.
- Wait for the email. OpenAI says exports can take up to 7 days to arrive, which is the regulatory ceiling rather than a realistic wait, though plenty of people report a few days. The link then expires 24 hours after you receive it, so the deadline that actually bites is the second one.
- Download the ZIP and extract it somewhere sensible rather than leaving it in Downloads.
Pro tip
Most people read the old advice about this export, which was to install a browser extension or find somebody to write you a Python script. You do not need any of that. OpenAI ships a viewer inside the ZIP and it works offline. The script is only for turning the archive into something else, like Markdown or a PDF, and that is optional.
After the download
- Open chat.html first. It is the readable one, it needs nothing installed, and on a long history it is a large file that takes a while to load. It has to stay in the same folder as conversations.json, because it reads that file rather than containing the conversations itself.
- conversations.json is a tree, not a transcript. When you regenerate an answer, ChatGPT keeps the version you threw away as a branch next to the one you kept, and the JSON holds all of it. That is good for an archive and awkward if you want a clean read, so any converter you write has to walk the tree and pick a path.
- A folder per conversation does not exist. Everything arrives as one file, which is why chat.html is one enormous document rather than a directory you can browse. If you want individual files, a converter is the only route.
- The images come out in their own folders, generated ones separated by model, and they are the only part of this export you can open without thinking about it first.
- Nothing imports back. You can move a history to a new personal account by uploading the JSON into a chat, but Team and Enterprise workspaces are a different story, and deleting the workspace deletes the chats in it. Take the export before you do anything drastic with the account.
- Store the archive in ProtonDrive or pCloud for an off-site backup