If a conversation with an AI may ever matter — to a lawyer, a clinician, an insurer, or a court — the export is the first thing to do and the easiest thing to put off.
Do it today. Here is why the timing matters, and exactly how the ChatGPT export data process works so the file holds up later.
Key Takeaways
- The export usually arrives within hours but can take up to seven days, and the download link expires 24 hours after you receive it.
- The ZIP is account-wide and includes conversations.json with per-message timestamps and the model used for each answer.
- Never edit the original file, never delete the account, and do not go back into the thread.
- Keep two copies in two places, and request your medical records for the same window.
- If you searched how to download ChatGPT conversation history, the export below is the answer; knowing how to export ChatGPT data is only half of it, and the file must then be preserved without edits.
Two deadlines people miss
The export is not instant. OpenAI processes the request and sends a link by email, or by SMS if the account uses only a phone number. It usually arrives within hours, but OpenAI says it can take up to seven days. The ChatGPT export data request is queued, not immediate.
The link expires in 24 hours. Once the email arrives, you have one day to download the file. Miss it and you request the export again and start over — another wait, another window.
So with ChatGPT export data, the request and the download are two separate acts of remembering. Request it now; watch for the email, including in spam and promotions, from noreply@tm.openai.com.
How to export ChatGPT data: step by step
- Sign in at chatgpt.com on a computer with the account that holds the conversations. (The mobile app can do this too, but the file is easier to handle on a computer.)
- Click your name or profile picture, then Settings, then Data controls, then Export data. Confirm the export.
- Watch that account’s email for a message from OpenAI saying the data export is ready. It usually arrives within hours; OpenAI says it can take up to seven days. Click the download link. The link expires 24 hours after it is sent; if it lapses, request the export again.
- Save the ZIP file. Inside are conversations.json (every conversation, with the model recorded on each message), chat.html (a readable copy), user.json, message_feedback.json, and shared_conversations.json. Do not edit them.
- Screenshot Settings → Subscription (plan and billing history) and, under Settings → General, any setting for legacy models. Screenshot the model picker at the top of a conversation.
- Archived conversations are included. Conversations the user deleted are not. Temporary chats were never saved.
- If OpenAI deactivated the account, a data request goes through privacy.openai.com or dsar@openai.com; an authorized representative (POA, estate) emails dsar@openai.com with proof of authority.
- The export is how the model question gets answered. Nothing the chatbot says about itself substitutes for it.
What you actually get
The ChatGPT data export is account-wide. You cannot select a single conversation; it is all or nothing, which for evidentiary purposes is a feature rather than a limitation.
Archived conversations are included. Deleted conversations are not, and an export cannot restore them. Temporary chats were never saved to your history, so they are not in the export either.
Inside the ChatGPT data export:
- conversations.json — the complete machine-readable record. This is the important file. It carries per-message metadata, including timestamps for each turn and which model produced each response.
- chat.html — a browsable viewer. Convenient for reading, but treat the JSON as the record.
- Additional account files depending on your settings.user.json, message_feedback.json, and shared_conversations.json, plus any other account files OpenAI includes. Keep them all, unedited.
Per-message timestamps and per-turn model identification are what make the export worth far more than screenshots, and what make ChatGPT chat history evidence rather than recollection. A delayed-care or medication claim usually turns on sequence: what you described, when, what the answer was, and what you did next. The JSON establishes the first three without anyone reconstructing them from memory.
Preserving it properly
A file that exists is not the same as a file that will hold up. Once the ChatGPT export data lands, a few habits cost nothing now and matter later.
- Do not edit the export. Not the JSON, not the HTML. Keep the original .zip exactly as downloaded.
- Keep a second copy somewhere separate — cloud storage plus an external drive, or email it to yourself.
- Note the download date in a plain text file alongside it.
- Screenshot the key exchanges too. Redundancy is cheap, and screenshots are easier to show someone quickly. Screenshots do not save ChatGPT conversation metadata, so they supplement the export rather than replace it.
- Do not delete the conversation from your account, and do not delete the account. The export is a copy; the account is the source.
- Do not go back and continue the thread to “clarify” what it meant. New messages join the same record, and a conversation that ends at the injury reads very differently from one that continues afterward.
That last point deserves emphasis. The instinct to return to the chat and ask the AI why it said what it said is a strong one. Resist it. Whatever it answers is not evidence of what it was thinking — models have no memory of their reasoning — and the follow-up becomes part of the file a defense lawyer will read.
Other products
ChatGPT Health conversations sit within the same account and come through the standard export; you do not need to export ChatGPT conversations separately for Health, but health data linked from patient portals, labs, or wellness apps may have its own controls. Connected sources are listed under Health → Accounts. Do not disconnect them: OpenAI deletes the data synced from a source within 30 days of a disconnect. Request the underlying records directly from the provider portal that fed it.
Character.AI, Replika, and companion apps each have their own export path, usually under account settings or privacy. If an app offers no export, screenshot systematically — full screens, with dates visible, in order — and record the app version.
Google Gemini exports through Google Takeout. Claude exports through Settings → Privacy. Copilot runs through your Microsoft account privacy dashboard.
If you cannot find an export function, screenshot first and ask questions later. The steps to export ChatGPT conversations above are specific to OpenAI; other platforms differ. The export can be figured out; a deleted conversation cannot.
If someone in your family died
Account access after a death is genuinely difficult, and every week of delay narrows the options.
Do not attempt to log in and delete anything. Preserve the device itself — powered off if possible, and not wiped or reset. Keep any phone or computer the person used for the conversations. Then talk to a lawyer early, because the mechanisms for obtaining account records from a platform after a death are legal ones, and some of them are time-sensitive.
An authorized representative, such as an attorney-in-fact or the estate’s personal representative, can request the account’s data from OpenAI by email to dsar@openai.com with proof of authority to act.
If the conversation may be part of a claim
The export is step one. Knowing how to download ChatGPT conversation history is the easy part; keeping it intact is the rest. A reviewing attorney will also want the medical record for the same window, a timeline of what you did and when, and the details of which product and mode you were using.
The Schenk Law Firm reviews AI-related injury claims — medical advice and misdiagnosis and AI-related mental health harm. Review costs nothing and carries no obligation.
Why this firm
The Schenk Law Firm is a San Diego personal injury and mass tort firm that reviews AI medical advice claims nationwide. Attorney Benjamin S. Schenk was appointed by Hon. Ethan P. Schulman to serve as a member of the Plaintiffs’ Steering Committee, a leadership role in the California coordinated proceeding covering ChatGPT product liability cases. The Schenk Law Firm has recovered more than $25 billion on behalf of clients. Past results do not guarantee future outcomes.
We are experienced with emerging-technology injury claims — underage online gambling, social media addiction, and AI-related mental health litigation among them. Cases involving AI medical advice are new, the law governing them is developing, and no lawyer can promise an outcome in any of them. What we can tell you is whether a record shows the pattern a ChatGPT lawsuit requires.
The Schenk Law Firm, LLP · (858) 424-4444
General information, not legal advice. Export procedures are set by the platforms and change without notice; the steps above reflect what OpenAI documents as of September 2026.
