What You Get When You Export ChatGPT Data
When you request a data export from ChatGPT, OpenAI emails you a ZIP file containing everything associated with your account. Most people expect to find their conversations and not much else — but the export is actually more comprehensive than that, and understanding what's inside helps you make better use of it.
The ZIP typically arrives within a few minutes of requesting it, though accounts with very large histories may take longer. Once downloaded, you can extract it like any other ZIP file to reveal a collection of JSON files, HTML files, and any media you've generated.
The Files Inside the ZIP
conversations.json
This is the main file and almost certainly what you came for. It contains your entire ChatGPT conversation history in JSON format — every message you've sent and every response you've received, across every conversation, in chronological order.
Each conversation in the file includes:
- A unique conversation ID
- The conversation title (auto-generated by ChatGPT)
- The date and time it was created and last updated
- Every message in the thread, with timestamps and role labels (user or assistant)
- Any tool use, such as code interpreter runs or image generation requests
- Model information — which version of ChatGPT responded to each message
For heavy users, this file can be very large. It's not unusual for people with a year or two of regular ChatGPT use to have a conversations.json running to tens or even hundreds of megabytes.
message_feedback.json
This file logs any feedback you gave to ChatGPT responses — thumbs up, thumbs down, or written feedback. If you've never used the feedback buttons, this file will be mostly empty.
model_comparisons.json
If you ever participated in ChatGPT's model comparison feature (where two responses are shown side by side and you pick the better one), those comparisons are logged here.
user.json
Basic account information: your name, email address, account creation date, and a few other profile details. Nothing surprising here.
chat.html
A human-readable HTML version of your conversation history. You can open this directly in a browser and read through your chats without needing to parse any JSON. The formatting is basic but functional — useful if you just want to read through your history without setting up any tools.
dalle (folder)
If you've used ChatGPT to generate images with DALL-E, this folder contains all of them as image files. Depending on how much you've used the image generation feature, this can be a substantial portion of the overall ZIP size.
tts (folder)
Any text-to-speech audio generated during your conversations is stored here as audio files. Most users will find this folder empty or absent.
What the JSON Actually Looks Like
If you open conversations.json in a text editor, you'll see something that's technically readable but not particularly pleasant to work with. The structure is deeply nested — each conversation contains a mapping of message nodes, each of which contains content, metadata, and references to parent and child messages. It reflects the branching nature of ChatGPT conversations, where regenerating a response creates a new branch rather than overwriting the old one.
This means your export may contain more content than you expect — including responses you regenerated and replaced, which are still stored in the JSON even though they're no longer visible in the ChatGPT interface.
What's Not Included
A few things worth knowing about what the export doesn't contain:
- Files you uploaded — documents, images, or data files you sent to ChatGPT are not included in the export
- Custom GPT conversations — chats with GPTs you've built or used may not be fully included depending on your account type
- Billing information — payment details are not part of the data export
- Real-time browsing results — the web content ChatGPT retrieved during browsing sessions isn't stored, only the conversation text
Making the Export Actually Useful
The raw JSON is comprehensive but not practical. Reading through it manually, or trying to search it with a text editor, quickly becomes unwieldy for anything beyond the smallest exports.
The most straightforward way to work with a ChatGPT export is to load it into Export Reader, which is built specifically for this purpose. Upload the ZIP file directly — no extraction needed — and it handles the parsing automatically. You get a fully searchable archive of every conversation, with keyword highlighting, tag filtering, and the ability to save useful snippets. DALL-E images are extracted and browsable separately, and the AI Arranger can automatically group and summarise your conversations by topic.
Whether you're trying to find a specific piece of code, recover a detailed plan, or simply understand how you've been using ChatGPT over time, having the export loaded and searchable makes all of that practical rather than theoretical.
Conclusion
A ChatGPT export ZIP is more than just a copy of your chats — it's a structured snapshot of your entire account history, including images, audio, feedback, and metadata that the standard interface never surfaces. The conversations.json file alone can represent years of useful work, ideas, and problem-solving. The challenge is making it accessible, which is where having the right tool to parse and search it makes all the difference.