Why the paste breaks
Chat output is Markdown rendered by the chat UI. When you copy the rendered text, the browser hands over styled HTML that Word interprets loosely — heading levels collapse, nested lists lose their indentation, and a table arrives as a row of tab-separated words.
JotDoc reads what you paste as what it is: Markdown, HTML, or rich text from another editor. It parses the structure first and then renders it in the editor, so nothing has to be repaired by hand afterwards.
- Heading levels survive — including the ones ChatGPT writes as bold lines.
- Tables become tables, with header rows where the source had them.
- Code blocks keep their language and their line breaks.
- LaTeX left as $…$ or $$…$$ becomes a formula instead of dollar signs.
The second half: it is a draft, not a deliverable
Chat answers are plausible first drafts. The work that makes them sendable — checking a claim, tightening a paragraph, asking a colleague whether the argument holds — happens after the paste, and that is where a chat window stops being useful: every round means pasting the whole thing back in.
A JotDoc document stays put. Connect your AI app once over MCP and it reads the current text instead of your pasted copy, proposing changes you accept one at a time. Send a review link and a person can comment on the sentence they mean, without an account.
Three ways in, pick the one that matches
All three end at the same Export button, so the choice is only about how much you want the AI to touch the document afterwards.
- Paste — fastest. Good when the draft is basically right and you only need format and polish.
- Paste, then Hand off — copy one sentence from JotDoc into your own chat and paste the reply back as suggestions. No connection, no install.
- Connect over MCP — the AI works on the document directly: Claude, ChatGPT connectors, Claude Code, Cursor. Changes still arrive as suggestions.
The comparison, without the marketing
| Straight into Word | A chat canvas | JotDoc | |
|---|---|---|---|
| Structure after pasting | Flattened | Held in the chat | Kept as document structure |
| Ask for another revision | Re-paste the whole draft | Same thread, draft drifts | Hand off the paragraph |
| Someone else comments | Email the file around | Share the chat | One review link, no account |
| Word export | By hand | Copy out again | Export → Word |
Questions this page gets
Does this work with answers from Claude, Gemini and Copilot too?
Yes. The paste parser does not care which chat window the text came from — Markdown, rich text and plain text all land as structured blocks. There is a separate page for Claude if you want the specifics.
Do I need to install anything?
No. Pasting and exporting work in the browser with no account. A connector is only needed if you want the AI to read the document directly instead of you pasting each round.
Will long answers be truncated?
No practical limit on paste. Very long documents are handled as ordinary documents — the editor keeps up, and the browser stores your work locally as you type.
Where does my draft live?
In your browser by default, in local storage. It goes to the server only when you choose to make a cloud copy, share a review link or connect an AI app.
Where to go next
- AI apps over MCP — how the AI reads the latest text and why it can only suggest changes.
- Peer review without accounts — the step between “finished” and “sent”.
- Connect an AI app — five minutes, one endpoint.
- Working on a Mac? The desktop app opens the same documents from Finder.