Try it in thirty seconds

The demo archive is large on purpose, because that is when a reader stops being a nicety. Four things worth opening:

Search across every message — dozens of hits with the matching sentence under each one. The full text index is fetched only once you start typing.
Open a long thread — nearly fifty messages, with a question index down the right side.
See rendered maths — LaTeX drawn by KaTeX bundled locally, not fetched from a CDN.
Read the missing-files report — what the export referenced, what it actually shipped, and which conversations lost the most.

Your export arrived as conversations-000.json

If your history is large, OpenAI no longer ships a single conversations.json. It arrives split into conversations-000.json, conversations-001.json and so on, sometimes inside a nested ZIP. Viewers that look for one exact filename quietly show you a fraction of your history, or nothing at all.

This tool finds every shard through export_manifest.json, falls back to matching the filename pattern, and opens the inner archive that a Privacy Portal export wraps them in. Each file is streamed, so a 300 MB single-line JSON never lands in memory whole.

What the ZIP does not tell you

chat.html dumps every conversation onto one page with no search.
Attachments are named file_0a1b2c3d.dat with no extension, so nothing opens them. Extensions get restored from file signatures.
Exports routinely reference images they do not contain. You get a report of what is missing and which conversations lost the most.
Conversations are trees, not lists. The thread is rebuilt from current_node back through parent links, so edits and regenerations do not scramble the order.

Full-text search across a ChatGPT archive with the matching sentence highlighted under each result — Search runs over everything you and the model wrote, with the matching sentence under each hit.

Run it

Python 3.8 or newer, no dependencies.

git clone https://github.com/Gusarovv/chatgpt-export-viewer.git
cd chatgpt-export-viewer
python3 build_site.py ~/Downloads/your-export.zip

The site lands next to the export. Open index.html and read your archive the way you read a website. Copy the folder to a USB stick and it still works.

A ChatGPT conversation rendered as a static page with an inline image and collapsed reasoning steps — Reasoning and tool steps fold into one collapsible block per turn instead of being dropped.