Export your ChatGPT data, then convert conversations.json to Markdown. Each conversation becomes a .md file with speaker labels, fenced code blocks, and preserved formatting, ready to drop into Obsidian, Notion, Logseq, or any static site generator.
Short answer
Export your ChatGPT data, then convert conversations.json to Markdown. Each conversation becomes a .md file with speaker labels, fenced code blocks, and preserved formatting, ready to drop into Obsidian, Notion, Logseq, or any static site generator.
Markdown is the best default format for archived AI conversations. It is plain text, so it will open in anything forever, but it retains enough structure that headings, lists, code, and emphasis all survive. For anyone keeping a personal knowledge base, it is the obvious target.
Why Markdown suits AI conversations specifically
ChatGPT already writes in Markdown. Ask it for a comparison and you get a table; ask for code and you get a fenced block; ask for steps and you get a numbered list. The web interface renders that Markdown into HTML for display.
Converting to Markdown therefore does not translate anything, it recovers the original. Converting to PDF or Word is the lossy direction, because structure gets flattened into visual styling. Markdown keeps the semantics.
What good Markdown output looks like
The conversation title as an H1, a date line, speaker turns as H2 headings so they show up in outline views, and fenced code blocks with their language tag intact so syntax highlighting works in your editor.
Speaker turns as headings matters more than it sounds. In Obsidian and Logseq, that gives you a navigable outline of any long conversation in the sidebar.
Request your export from ChatGPT under Settings, then Data Controls, then Export Data
Drop the ZIP or conversations.json onto the converter
Select conversations, or all of them
Choose Markdown and download
Each conversation becomes a separate .md file named from its title, which is exactly the shape a notes vault wants. Bulk conversion turns an entire history into a folder you can drop straight into Obsidian.
Processing runs locally in your browser, so the export is never uploaded.
Getting it into Obsidian
Create a folder in your vault, drop the .md files in, and Obsidian indexes them immediately. They become searchable alongside your own notes, you can link to them with wikilinks, and graph view will show them.
A tip worth applying: add a consistent tag or put them all under one parent folder, so you can distinguish AI-generated content from your own writing when searching. Conversations tend to be verbose, and without separation they can crowd out your own notes in search results.
Getting it into Notion
Notion imports Markdown files directly. Use Import, then Markdown, and select the files. Notion converts each into a page, preserving headings, lists, and code blocks. Tables survive too, provided the converter emitted proper Markdown table syntax rather than ASCII art.
Filename handling
Conversation titles frequently contain characters that filesystems reject, such as slashes and colons, and ChatGPT auto-generates titles that can be long. A converter has to sanitise these without producing collisions. ChatExports strips illegal characters, truncates sensibly, and appends a short discriminator when two conversations share a title.
Frequently asked questions
Does Markdown conversion keep code blocks?
Yes, as fenced blocks with the language tag preserved, so syntax highlighting works in your editor.
Can I import the Markdown files into Obsidian?
Yes. Drop the files into any folder in your vault and Obsidian indexes them straight away, with no import step required.
Do tables survive the conversion?
Yes. Tables ChatGPT produced are emitted as proper Markdown table syntax rather than plain text, so they render correctly in any Markdown viewer.
Is Markdown better than PDF for archiving?
For durability and searchability, yes. Markdown is plain text and will open in anything indefinitely. PDF is better only when the document must be a fixed, unalterable record.