How do I access the ChatGPT archive?

In ChatGPT, open Settings → Data Controls → Export Data. OpenAI emails a ZIP with conversations.json containing every message from your account.

How do I export my Claude conversation history?

In Claude, open Settings → Privacy → Export Data. Anthropic emails a link to a JSON archive of every conversation on your account.

How do I find archived chats after exporting?

Import the exported JSON into a local-first tool that indexes each message. Semantic search then surfaces chats by meaning, not just exact keywords — so a search for 'kubernetes rollout' finds threads that only said 'k8s deploy'.

← Home

How to Export & Search Your AI Chatbot Conversations Archive

Your ChatGPT and Claude threads are one of the most valuable knowledge bases you own — decisions, drafts, debugging sessions, research. This guide covers how to export that history, how to access archived chats, and how to actually search a personal AI archive without handing it back to a third-party cloud.

Why export your AI history at all?

Provider access can change. Accounts get locked, plans get downgraded, retention policies shift. An export is your copy.
Built-in search is shallow. ChatGPT and Claude match on titles and recent messages, not the full body of every thread you've ever had.
Privacy. A local archive keeps sensitive conversations — health, finance, client work — off someone else's server.
Cross-tool memory. Once your ChatGPT and Claude history live in one place, you can search across them together.

How to access the ChatGPT archive

Open ChatGPT and click your profile → Settings.
Go to Data Controls.
Click Export data and confirm.
OpenAI emails a download link (usually within an hour). The ZIP contains conversations.json with every message on your account, plus chat.html for a readable view.

The link expires in about 24 hours — download it promptly and store the ZIP somewhere you control.

How to export Claude conversation history

Open Claude and go to Settings → Privacy.
Click Export data.
Anthropic emails a link to a JSON archive containing every conversation and message on your account.

Common use cases for an AI history archive

Finding archived chats by meaning, not title. "That thread where we figured out the auth bug" — semantic search finds it even if the title is Untitled.
Rebuilding context on a project. Pull every conversation touching a client, repo, or product line into one view.
Reusing prompts that worked. Turn one-off prompts you actually got value from into a template library.
Auditing what you shared. Search your archive for API keys, addresses, or client names you never meant to paste.
Personal research memory. Long-running learning threads (a language, a framework, a domain) stay searchable across months instead of scrolling a sidebar.
Handover and continuity. Export before a job change or account migration so the knowledge moves with you.

Why local-first beats cloud archiving

Most "AI history" tools re-upload your export to their own servers and index it there. That trades one silo for another — and often a less-audited one.

A local-first archive keeps the raw export and its search index on your device. The provider you exported from no longer sees your queries, and neither does the archive tool. That matters most for the exact conversations you were nervous about having in the cloud to begin with.

Semantic search vs. keyword search on personal archives

Keyword search on conversations.json only finds threads that literally contain the words you type. Personal archives don't work that way — you remember the shape of a conversation, not its exact vocabulary.

Semantic search embeds each message into a vector and matches by meaning. A search for "how we priced the enterprise tier"surfaces the thread that actually said "annual contract math for the big customer" — no shared keywords, same idea. On a multi-year archive this is the difference between finding the thread in seconds and never finding it at all.

Putting it together with Marginalia

Marginalia is built for exactly this workflow: import your ChatGPT and Claude exports, keep the index on your device, and search by meaning across every thread you've ever had. No re-upload, no server-side history, no lock-in — just your AI conversations, finally searchable.

Try it on your own archive →

FAQ

How do I find archived chats in ChatGPT?

ChatGPT doesn't have a separate archive folder for old threads — "archived" just means older than the sidebar shows. The reliable way to find them is to export your data and search the export.

Does exporting delete my chats from the provider?

No. Export is a copy. Deletion is a separate action in the same settings area.

Can I export a single conversation instead of everything?

Both ChatGPT and Claude only support account-wide exports today. Filter after import.