How long does export take?

Up to 7 days as of June 2026 (OpenAI's own help text), though many land within hours. Do not assume failure before 7 days.

What sender does the export email come from?

`noreply@tm.openai.com`. Older guides say plain `openai.com`; both belong to OpenAI, so filter on `openai.com` to catch either.

Most often a Business/Enterprise workspace where self-serve export does not run, or you opened it in the wrong account identity.

Why won't my ZIP unzip?

A large export truncated mid-download. Re-request a fresh link, download on a stable connection, and try a different browser or `7-Zip`/`The Unarchiver`.

Can I cancel an in-flight export?

No, but submitting a new request automatically voids the old one, so only your latest request stays live.

What's in the export?

Conversations (`conversations.json`), a browsable `chat.html`, account/user files, and your Memory/custom-instructions data. Recently uploaded files are included for a limited window; not all generated images are.

Nothing works — what's the last resort?

Submit a formal data-access request at [privacyportal.openai.com](https://privacyportal.openai.com) under GDPR or CCPA; OpenAI responds within ~30 days, independent of the in-app export.

Troubleshooting

ChatGPT Export Conversations Failed: Fix the Missing or Empty ZIP

Your Data Controls export never emails a ZIP, the ZIP is empty, or it downloads corrupted. Find which link in the chain broke and recover your full archive.

Published: May 24, 2026 Updated: Jun 15, 2026 Author: AI Productivity Guide Team 🌐 查看中文版本

You opened Settings, then Data Controls, clicked Export, saw “Your data is being prepared for download”, and then nothing landed. Or the email arrived but the ZIP is suspiciously small with one HTML file and no real conversations. Or the ZIP downloads but your unzip tool throws End-of-central-directory signature not found. ChatGPT’s export runs as an async background job, and any link in the chain (queue, email, signed download URL, archive integrity, account scope) can fail silently.

Fastest fix: search your whole mailbox (including Spam and Promotions) for from:noreply@tm.openai.com first, because as of June 2026 OpenAI sends the export link from the tm.openai.com subdomain and Gmail loves to file it under Promotions. If nothing is there after a day, re-request from the exact login identity you signed up with, and use a personal email if your corporate filter is eating it. The export now takes up to 7 days to arrive (not minutes), so do not assume failure too early.

Which bucket are you in?

Symptom	Most likely cause	Jump to
No email after hours/days	Spam/Promotions, wrong sender filter, or queue still running (up to 7 days)	Cause 1, 2
Email link gives an error	Signed link expired (24h) or a newer request voided it	Cause 3
ZIP downloads but is tiny/empty	Wrong account identity, or Business/Enterprise (chat export unavailable)	Cause 4, 5
ZIP downloads but won’t unzip	Truncated/corrupted download of a large archive	Cause 6
Brand-new account, empty ZIP	Data not yet propagated to the export pipeline	Cause 7

Common causes

Ordered by hit rate, highest first.

1. Export email landed in spam, Promotions, or was filtered by sender

As of June 2026 OpenAI sends the export link from noreply@tm.openai.com (the tm.openai.com subdomain, which it also uses for workspace invites). Older guides tell you to search for plain openai.com, so people filter on the wrong sender and miss it. Corporate spam filters, Gmail’s Promotions tab, or strict DMARC on your domain can hide or reject it.

How to judge: Search your mailbox across all folders (Spam, Trash, Promotions) for from:openai.com — that wildcard catches both tm.openai.com and any legacy address. No hit anywhere after 7 days means the queue is still running or the message was rejected upstream.

2. Export job still running in the queue

When usage spikes, the background job queue backs up. The UI shows “preparing” and the email has not sent yet. OpenAI’s own help text now says exports can take up to 7 days (as of June 2026, up from the old “minutes to a few hours”), so a 6-hour wait is not a failure.

How to judge: Fewer than 7 days since you clicked Export and still no email is normal — wait. Past 7 days with nothing in any folder means the job stalled; re-request to reset it.

3. Download URL expired, or a newer request voided it

The link in the email is signed and short-lived: 24 hours from when it was sent. There is a second, less obvious trap — OpenAI only fulfils your most recent export request. Every time you click Export again, all earlier links are silently cancelled, even ones that have not expired yet.

How to judge: Click the link. Link expired, Access denied, or a 404 means it is dead. If you requested export several times, only the email from your last request is live; the rest are dead on arrival.

4. Scope mismatch with the wrong account identity

If you signed up via Google originally but requested the export while logged in with email/password (or via Apple, Microsoft, or SSO), the export covers only the identity you are currently in. The other account’s chats are not included.

How to judge: Open Settings, then Account, and compare the email shown there to the one you used at signup. A mismatch means you exported a different account than the one holding your chats. Note also: phone-number-only accounts receive the export link by SMS, not email.

5. Business or Enterprise account — chat export is unavailable by design

This is the most common reason for an “empty” or metadata-only ZIP. As of June 2026, the self-serve chat export in Data Controls is not available for ChatGPT Business or Enterprise workspaces. The Export button may be hidden, or it returns only account metadata, not conversations.

How to judge: Settings, then Account, shows your plan. If it reads Business or Enterprise, the personal chat export simply does not run there. Only a workspace admin can pull workspace data through the compliance/admin tooling, and an admin export is governed by your org’s data-retention policy.

6. ZIP downloaded but is corrupted or truncated

Large exports (multi-GB, e.g. heavy image-generation history) sometimes truncate mid-download. The signed URL is real, the file just arrives incomplete. Reported repeatedly in June 2026.

How to judge: Your unzip tool errors out. Typical strings:

macOS Archive Utility: Unable to expand "<file>.zip"
The Unarchiver: Could not extract <file>.dat
Terminal unzip: End-of-central-directory signature not found ... cannot find zipfile directory

Re-download the same link twice and compare byte counts (right-click, Get Info / Properties). If the two downloads differ in size, the transfer is truncating, not the source file. See the fix in Step 6.

7. Account too new or has too few conversations

Brand-new accounts can produce empty exports because the data has not finished propagating to the export pipeline.

How to judge: Was the account created less than 24 hours ago, or do you have only one or two chats? Wait a day and re-request.

Before you start

Confirm the email address (or phone number) on your ChatGPT account — that is where the export link goes.
Decide if you need everything or just a few chats. Sharing individual chats is faster for small needs.
If you are on Business or Enterprise, know that personal chat export will not run — go to your admin or the Share fallback.

Information to collect

Account login method (email, Google, Microsoft, Apple, SSO, or phone-only).
The exact email address or phone number tied to that login.
Approximate date of your most recent conversation — the ZIP should reach at least that date.
Plan tier and, for Business/Enterprise, your workspace name.
Whether the email/SMS arrived at all and, if so, the timestamp.
Browser and connection used for the download (relevant for truncated ZIPs).

Step-by-step fix

Step 1: Search every mail folder for the export message

In Gmail, search across all mail and spam:

from:openai.com subject:"data export"

The bare openai.com filter catches the current noreply@tm.openai.com sender and any legacy address. Check Spam, Trash, and Promotions explicitly. If found, click the link within 24 hours of its timestamp. On a phone-only account, check your SMS messages instead.

Step 2: Request export again from the right identity

If nothing turns up, sign out completely, then sign back in with the exact method you originally used (Google vs email vs Apple/SSO matters). Go to Settings, then Data Controls, then under Export data click Export, then Confirm export. Watch for the confirmation banner. Request only once and wait — a second request voids this one.

Step 3: Allowlist OpenAI so the next email is not filtered

In Gmail: Settings, then Filters and Blocked Addresses, then Create a new filter, with from:openai.com, then check “Never send it to Spam”. This covers tm.openai.com too and stops future exports from being silently filtered.

Step 4: For corporate inboxes, switch to a personal email

If your work email blocks the message (strict DMARC, gateway quarantine), change your ChatGPT account email to a personal Gmail under Settings, then Account, wait for the verification email, then request the export again so the link lands at the new address.

Step 5: If the ZIP is empty and you are on Business/Enterprise

Self-serve chat export does not run on those workspaces. Ask your workspace admin to pull the data through the admin/compliance tooling, governed by your org’s retention policy. Send them the timestamp of your failed attempt and a screenshot of the empty ZIP.

Step 6: For a corrupted or truncated ZIP, re-download cleanly

Request a fresh export so you have a new, full-validity link (the old one may already be near its 24h limit).
Download over a stable wired or strong Wi-Fi connection — truncation usually comes from a dropped transfer on a multi-GB file.
Try a different browser (Chrome vs Safari/Firefox); some users only get a complete file on the second browser.
Unzip with 7-Zip (Windows) or The Unarchiver (macOS) rather than the OS default, and run a terminal check: unzip -t export.zip reports No errors detected on a good archive.
Compare the downloaded byte size between two attempts; if they differ, the download truncated and you should retry, not “repair”.

If you only need a few chats and export is broken, open each one, click Share, then copy link, and save the page as HTML or PDF from your browser’s print dialog. Less clean than a full export, but reliable.

Step 8: If everything fails, use the Privacy Portal (GDPR/CCPA)

The in-app export and the formal data-access request run on separate systems. Go to privacyportal.openai.com and submit a data access request under GDPR (EU/UK) or CCPA (California). OpenAI typically responds within 30 days, and this path works even when the in-app Export button is broken. You can also open a ticket at help.openai.com — include your account email, the timestamp of the failed request, and a Data Controls screenshot.

How to confirm it’s fixed

The email arrives from noreply@tm.openai.com (or SMS, for phone accounts).
The ZIP downloads without error and is well over a few KB.
unzip -t export.zip reports No errors detected (or your unzip tool opens it cleanly).
Inside the ZIP, chat.html opens in a browser and shows your full chat list; conversations.json is present and non-empty.
Spot-check your most recent conversation — it should be present and complete up to at least its real date.

Long-term prevention

Request an export quarterly as a personal backup, not just when something breaks.
Keep a personal email (one you actively monitor) on the account, and pre-allowlist openai.com.
Save important conversations as PDFs when you have them, instead of relying solely on export.
On Business/Enterprise, confirm with your admin how to retrieve data before you need it — self-serve export will not be there.
After every successful export, verify the ZIP opens (unzip -t) before deleting the email.

Common pitfalls

Re-requesting export 5 times in a row — only the latest link is valid, all earlier emails are now dead.
Clicking an expired (24h) link and assuming export failed — it just expired, request again.
Searching only the Inbox, not Spam or Promotions, and filtering on the wrong sender domain.
Exporting from the wrong identity and concluding “ChatGPT has no data on me”.
Expecting Business/Enterprise self-serve export to work — it does not; that path runs through an admin.
“Repairing” a truncated archive instead of re-downloading the file on a stable connection.

FAQ

How long does export take? Up to 7 days as of June 2026 (OpenAI’s own help text), though many land within hours. Do not assume failure before 7 days.
What sender does the export email come from? noreply@tm.openai.com. Older guides say plain openai.com; both belong to OpenAI, so filter on openai.com to catch either.
Why is my ZIP empty? Most often a Business/Enterprise workspace where self-serve export does not run, or you opened it in the wrong account identity.
Why won’t my ZIP unzip? A large export truncated mid-download. Re-request a fresh link, download on a stable connection, and try a different browser or 7-Zip/The Unarchiver.
Can I cancel an in-flight export? No, but submitting a new request automatically voids the old one, so only your latest request stays live.
What’s in the export? Conversations (conversations.json), a browsable chat.html, account/user files, and your Memory/custom-instructions data. Recently uploaded files are included for a limited window; not all generated images are.
Nothing works — what’s the last resort? Submit a formal data-access request at privacyportal.openai.com under GDPR or CCPA; OpenAI responds within ~30 days, independent of the in-app export.

Tags: #ChatGPT #Troubleshooting #account