How to import your ChatGPT chat history into Whizi

The short answer

Yes, Whizi imports ChatGPT conversation history. Export your data from ChatGPT, then open Settings, choose Import Chats under Account, and pick the conversations.json file out of the export. Whizi accepts exactly two import sources, ChatGPT and Claude, and the same dialog handles both.

The export file itself is never uploaded: your browser reads conversations.json or the raw export .zip, parses it, and posts only plain normalized conversations. The import is also safe to run more than once, because every imported conversation is keyed by account plus source plus source id, so importing the same export twice or clicking Import twice lands the same chats once.

Import failures surface as a small set of fixed messages, each listed with its trigger and its fix on chat import errors. Two of them are catch-alls: one for a file the browser could not read, one for an upload that did not finish.

What you need	Where it comes from
The conversations.json file	Inside the data export ChatGPT produces for you
The Import Chats control	Settings, under Account, described as a one-time setup action
A signed-in Whizi account	The import writes chats into the account you are signed in as

Picking the right file out of the export

The file Whizi reads is conversations.json. Everything else in a ChatGPT export archive is ignored, so the fastest route is to unzip the export and pick that file directly.

The .zip works too, and for a big history it is the better choice. A raw .json export is accepted up to 300 MB, past which you are told to use the .zip instead. The .zip path has no size ceiling of its own, because only the conversations.json entry is decompressed and the images in the archive never have to fit in memory.

You do not have to tell Whizi which product the export came from. The parser identifies the format itself, which is also how it knows to apply the Claude rules to a Claude export and the ChatGPT rules to this one.

What carries over

An import brings the transcript across, not the scaffolding around it. That is deliberate, and it is why a successful import can still look shorter than the original.

Two rules are specific to a ChatGPT export. Only the branch that was on screen is replayed, so regenerated replies and discarded drafts do not come across as messages. And non-transcript blocks are skipped: hidden reasoning, analysis tool code and output, browsing scratchpads, custom instruction blocks, and function calls.

The rest of the drop rules, and what they do to titles, long messages and odd timestamps, are listed on chat import errors.

A large export slices itself

The import endpoint takes conversations in batches rather than in one lump, and the website importer slices your export to fit, so a big history needs no splitting by hand and no setting changed. The per-request numbers, and the headroom the website leaves under them, are on chat import errors.

Whizi publishes no cap on the total number of conversations an account may import, and no per-day import cap. Alongside the 20 imports per minute route limit, the account-wide ceiling of 30 requests per minute and the per-IP ceiling of 60 per minute apply to every route, import included.

After the import

Imported conversations land in your chat list like any other Whizi chat. The sidebar search box finds them by title, and it matches the chat title only rather than message bodies.

On privacy, the import path is easy to reason about, because the export file never reaches a server in the first place. Once the conversations are stored they are treated like any other chat, and what Whizi does and does not do with them, training and retention included, is set out in what Whizi does with your data.