Export your Claude data under Settings, then Privacy, then Export Data, and drop the ZIP into ChatExports. Each conversation becomes a Word .docx with speaker headings, formatted code blocks, and artifacts included, named from the conversation title.
Short answer
Export your Claude data under Settings, then Privacy, then Export Data, and drop the ZIP into ChatExports. Each conversation becomes a Word .docx with speaker headings, formatted code blocks, and artifacts included, named from the conversation title.
Step by step
Export from Claude: Settings, then Privacy, then Export Data. Anthropic emails a link to a ZIP containing conversations.json.
Open the converter: Drop the ZIP onto ChatExports. Conversations are listed with dates and message counts.
Select and choose Word: Pick the conversations you want and select Word as the output format.
Download: Each conversation downloads as its own .docx, ready to open, edit, or send.
Copy and paste from Claude into Word looks fine for two paragraphs and falls apart on anything real. Code loses its monospace and indentation, long responses arrive as one undifferentiated block, and you lose track of who said what.
What a proper .docx contains
Speaker headings for each turn, so the document has structure the navigation pane can use
Code blocks in monospace with indentation intact and wrapping instead of clipping
Artifacts included as their own sections, since Claude puts substantial documents and code there
Timestamps on each message for records that need a date
Working lists and tables, converted to Word constructs rather than pasted characters
Full Unicode, so emoji and non-Latin text render instead of appearing as boxes
Why Word specifically
Word is where documents go to be worked on by other people. Tracked changes, comments, and review cycles all assume .docx, which is why a Claude thread destined for a client, a compliance file, or a colleague should be a document and not a screenshot.
Artifacts are the Claude-specific problem
Claude puts generated code, documents, and structured output into artifacts rather than the message body. A converter that only reads message text produces a document full of references to work that is not in it. Artifacts belong in the file, labelled and in position.
Long conversations
Claude threads run long, and extension-based scrapers see only the part of the thread the browser has rendered. Converting from the export avoids that entirely: the JSON holds every message regardless of what was on screen.
Batch conversion
Selecting many conversations produces one .docx each, named from the conversation title, so a project's worth of threads lands as a browsable folder rather than a single unmanageable file.
Convert Claude to Word in the browser. Nothing is uploaded, which matters when the threads contain client work or private code.
Frequently asked questions
Do code blocks survive the conversion?
Yes. They stay in monospace with indentation preserved and wrap rather than getting clipped at the page edge.
Are Claude artifacts included?
Yes. Artifacts are converted as their own labelled sections in the document.
Can I convert every conversation at once?
Yes. Select all and each conversation downloads as a separate .docx named from its title.
Does the file open in Google Docs or Pages?
Yes. The output is standard .docx, which Google Docs, Pages, and LibreOffice all open.