ChatGPT export search reader
Published . Updated .
Drop your ChatGPT data export into the reader below and it lists every web search query saved in your chats, with the pages each answer fetched and cited. It runs in this tab, so the file is never uploaded, and both lists download as CSV.
One catch: since 14 September 2026 ChatGPT no longer saves the query text in a chat, so newer chats show their pages and no queries. Why the queries went missing explains the change, and the reader tells you how many searches it hit.
The file is read in this tab. Nothing is uploaded.
How do you get your ChatGPT data export?
- Sign in to ChatGPT on the web and open your profile menu, then Settings.
- Open Data controls, start the export and confirm it.
- Wait for OpenAI to send the download link. OpenAI says the export can take up to seven days, and the link expires 24 hours after you get it.
- Drop the ZIP into the reader above. You do not need to unzip it first.
A large account can arrive with its chats split across several files named conversations-001.json and so on. Drop them all at once and the reader merges them.
What does the reader find in the export?
It finds the searches each answer ran and the pages it used, then lays them out as two tables. Every chat in the file is read, and a chat that appears twice is counted once.
| What | Where it sits in the export | How the reader shows it |
|---|---|---|
| Search queries | The search steps saved with each answer, including older browsing steps | Queries tab, one row per query, in the order ChatGPT ran them |
| Pages fetched | The search results saved with the answer | Pages tab, fetched marked yes |
| Pages cited | The answer's citations and sources, plus places and products that link a website | Pages tab, cited marked yes |
| Searches with no query text | Search steps that hold no query, which is every search since 14 September 2026 | Counted in a note above the tables |
A page that was fetched and cited is one row with two yeses, so you can see straight away which reads turned into citations.
Why do recent chats show pages but no queries?
Because ChatGPT stopped saving the query text. Since 14 September 2026 the searches reach your browser only in the live answer stream, so the saved chat, and the export built from it, keeps the search steps and the pages but not the words searched.
The reader counts those empty search steps and says how many answers have them, so a gap never looks like a chat that did not search. To keep the queries for chats from now on, the Only Fan Out side panel reads them while the answer is written and keeps them in your browser.
Is your export uploaded anywhere?
No. The reader is a script that runs in this tab: your browser opens the file, the tables are built in memory, and closing the tab clears them. The site's analytics count the visit and that the reader was used, never anything in the file.
What do the CSV columns mean?
Each CSV is one table, with one row per query or one row per page in each answer. If you set a filter, the download holds only the rows that match it.
| Column | File | Meaning |
|---|---|---|
| chat, date, turn | Both | The chat title, the day it started and which answer in the chat the row belongs to |
| prompt | Queries | Your message that started that answer |
| order, kind, query | Queries | Where the query came in the answer's searches, its kind (web, business, image or product) and the words searched |
| recency, domain, location | Queries | Settings ChatGPT sent with the query, such as a time limit, a site limit or a place, when saved |
| url, domain, title, kind | Pages | The page, its site, its title and whether it was a page, a place or a product |
| fetched, cited | Pages | Yes or no for each, so a page used both ways has two yeses |
Find steps, where ChatGPT looks for a phrase inside a page it already opened, are left out of the query list because they are not web searches. ChatGPT's kinds of fan-out query covers what each kind does.
What can you do with the two lists?
Use the queries to see what ChatGPT looks for on your topics, and the pages to see who it reads and who it cites. Sort the Pages CSV by domain and your own site's rows show you where you were fetched and not cited.
Our guide to using fan-out data for GEO turns both lists into a plan, and how to export ChatGPT search queries to CSV gives the spreadsheet layout and formulas we use.
Read next
- How do you export ChatGPT's search queries to CSV?
- Why ChatGPT's fan-out queries went missing in September 2026
- How to see what ChatGPT searched before it answered
- Why did ChatGPT read your page but not cite it?
- How do you use fan-out queries for GEO?
Sources
Related questions
Does the reader work with Claude or Grok exports?
No. It reads the ChatGPT export format only. For Claude and Grok, the Only Fan Out side panel lists the searches as you chat and exports them as CSV.
How large an export can it open?
It reads the whole file in your browser's memory, so a very large export can take a while. If a big ZIP will not open, unzip it and drop the conversations JSON files instead.
My export has several conversations files. Which one do I drop?
All of them, at once. The reader merges the files and counts a chat once if it appears in more than one.
Is the reader free?
Yes. There is no sign-up and no limit, and it keeps working without the extension.
Get the queries for new chats too
ChatGPT no longer saves its search queries, but they still reach your browser while the answer is written. Only Fan Out is a free Chrome side panel that keeps them, with the fetched and cited pages, for ChatGPT, Claude and Grok. Nothing leaves your browser.
Add to Chrome, it's free