Turn a ChatGPT chat into a spreadsheet.
A conversation is a list of messages in order, which is to say it is already a table. Extry writes it out as a .csv — one row per message, six columns.
- retry-backoff-design.csv9.8 KBtable
Why CSV
The format every spreadsheet already opens
CSV needs no importer and no add-on. Excel, Numbers, Sheets and every database loader have read it for decades, and the file starts with a byte-order mark so Excel does not garble accents or Cyrillic on the way in.
What it will not do is look like a conversation. A long answer sits in a single cell, and reading it there is worse than reading it almost anywhere else.
What the file looks like
A header row, then one row per message
Six columns: index, role, author, model, created_at and text. The header comes first, then the messages in the order they were sent.
A value containing a comma, a quote or a line break is wrapped in quotes, and quotes inside it are doubled. Those are the ordinary CSV rules, so anything that reads CSV reads this.
index,role,author,model,created_at,text
1,user,user,,2026-03-11T09:14:22Z,Why the hammering?
2,assistant,ChatGPT,gpt-5,2026-03-11T09:14:31Z,"Constant delay, no jitter."What you can do with it
Work a spreadsheet is good at
This is the part where a table beats a document.
Sort and filter
Show only what the assistant said, or only what came after a date, without scrolling the conversation.
Count and chart
Messages per model, per day, per role — a pivot table away.
Hand it to a tool
Database loaders, BI tools and annotation software take CSV as their default input.
How it works
Three steps, one file
- Open the conversation in ChatGPT and click the Extry button on the page.
- Pick CSV in the panel.
- Press Download this chat and open the .csv from the archive.
No account, no upload step. The table is built in the page, so the conversation is not sent to a converter.
Processing
Where each part runs
| What | Where it happens |
|---|---|
| Export to every format | In your browser |
| Attachments and images | In your browser |
| Conversation content | Not sent anywhere |
| Usage statistics | Not collected |
| Typeset PDF (later, opt-in) | On a server, file deleted right after |
Questions
About the CSV export
Which columns does it have?
Six: index, role, author, model, created_at and text. The header row names them, and role is the raw value — user or assistant — rather than the label the interface shows.
Will Excel garble non-Latin text?
No. The file starts with a byte-order mark, which is what tells Excel to read it as UTF-8.
What happens to a long answer?
It stays in one cell, quoted, with its line breaks intact. Fine in the formula bar, awkward in a column.
Does the code formatting survive?
The text keeps its Markdown, fences and all. A spreadsheet will not colour it, but nothing is thrown away.
More
Other shapes for the same chat
The same conversation, written differently: JSON, Markdown, Word, the main page.