ChatGPT Exporter — Save the Whole Chat You're Reading to Markdown or PDF

Open a long conversation in ChatGPT, scroll to the top, and count the messages. You will usually find fewer than were actually said. The page only keeps part of a long thread on screen — in a first-party test on 2026-08-17, a 44-message thread showed 10 messages when opened, and scrolling as far as the page would go only reached 31. Messages 15 through 27 were never drawn at all. That is the reason most "save this chat" tools quietly hand you half a conversation. This one does not read the screen. It asks for the conversation you have open, gets the full message list, and writes the whole thing to a file. WHAT IT DOES, IN ONE SENTENCE It saves the conversation you are currently reading as a Markdown file or a PDF — complete, in order, with the formatting intact. WHY "IT LOOKED FINE ON SCREEN" IS NOT THE SAME AS "IT SAVED FINE" Long threads are rendered a slice at a time. Whatever is off screen is not in the page, so anything that copies from the page copies a gap — and a gap is hard to notice, because the file you get still looks like a conversation. It just is not yours. This extension takes the other route. It asks for the single conversation you have open, using the session already in your browser, and works from the message list rather than from what happens to be painted. Long threads come out whole. Short threads come out identical either way; the difference only shows up where it matters. TWO FORMATS, AND A REASON FOR EACH Markdown — for notes, for a repository, for anything that has to stay searchable and diffable years from now. Headings, lists, tables, block quotes and fenced code blocks all survive as Markdown, so what lands in your editor has real structure instead of being a wall of text. It is an ordinary .md file on your disk: open it, move it, commit it, or drop it into whichever notes app you already use. PDF — for sending to someone, filing, or printing. It goes through your browser's own print engine, so what you get matches what you saw, and images from the conversation are embedded in the file itself. ONE DELIBERATE DIFFERENCE BETWEEN THE TWO Images are embedded in the PDF and left as placeholders in the Markdown. That is a decision, not a limitation. A test conversation with five pictures came to 1.5 MB per image; base64 encoding those into a Markdown file inflates them by a third and turns a small text file into roughly 9 MB. Markdown earns its place by opening anywhere, going into version control and being searchable — a 9 MB blob of encoded image data destroys all three. So the PDF carries the pictures and the Markdown stays a text file, and it tells you which images it left behind. WHAT DOES NOT END UP IN YOUR FILE Tool-call messages. When the assistant runs a search or a code tool, the raw internal call is a message in the thread but not part of what you were reading. An early build let those through and a 41-message conversation came out with 49 entries. They are filtered now, and the filtering is nailed down by tests — including the case where one message carries both a generated image and the internal text that produced it, where the picture is kept and the text is dropped. FAILURES TELL YOU WHAT HAPPENED If something goes wrong there is always a visible response, it is written in plain language, and it tells you what to try next. Every failure it knows how to hit — an expired session, rate limiting, a blocked print window, the daily allowance used up, even an error code it has never seen before — has a written message and a suggested next step, and tests hold them to that. An export that fails does not spend one of your daily exports. FREE AND PRO Free, without an account: - 2 exports per day - Both formats, complete output, no watermark, no truncation Free, signed in with Google: - 5 exports per day - Everything above, unchanged Pro — $2.99 per month, or $19.99 per year: - Unlimited exports - Identical output to the free tiers - Cancel at any time The paid plan lifts the daily count and nothing else. There is no format held back, no watermark on free exports, no cap on conversation length, and no "first N exports then it stops forever" — the allowance is per day and it comes back. Signing in is only ever needed to raise the free allowance or to carry a subscription between computers. WHAT IT ASKS FOR, AND WHAT IT NEVER TOUCHES - One permission: storage. It holds your settings, how many exports you have used today, a short-lived copy of your subscription status, and — only if you sign in — your sign-in token. - Your conversations never reach our servers. Reading, formatting and writing the file all happen inside your browser. There is no upload step, because there is nothing to upload to. - It only ever touches the one conversation you have open. It does not list your other conversations, does not walk your history, and cannot bulk-download your archive. That boundary is enforced in the code by a wrapper that refuses any request outside the current conversation, and it is covered by tests. - On chatgpt.com it reads only the conversation you have open. The only other site it can touch is our own account page at ext.mychatexport.com, and only to sign you in and to check whether you have a subscription — no conversation content is ever sent there. There is no third site. WHO IT IS FOR - People who use long threads as working documents — research, debugging, drafting — and need a copy that is not trapped in a browser tab. - Anyone building a personal knowledge base, who wants conversations in Markdown next to their other notes rather than in yet another web app. - Anyone who has to hand a conversation to someone else — a colleague, a client, a record — and needs a PDF that reads the way the conversation read. QUICK START 1. Install the extension. 2. Open any conversation on chatgpt.com. 3. Use the Markdown or PDF button on the page. 4. The file downloads, named after the conversation. There is nothing to configure and no account to cr…