Export your ChatGPT data, then convert conversations.json to .docx. Each conversation becomes an editable Word document with speaker labels as headings, formatted code blocks, and preserved lists and tables, generated locally in your browser.
Short answer
Export your ChatGPT data, then convert conversations.json to .docx. Each conversation becomes an editable Word document with speaker labels as headings, formatted code blocks, and preserved lists and tables, generated locally in your browser.
Word is the right target when a conversation needs to be worked on by people rather than machines: annotated, tracked, circulated for review, or dropped into an existing document template.
Why copy and paste falls short
Pasting from the browser into Word carries the web page's HTML styling with it. You get ChatGPT's fonts, its background shading, and its link colours, all of which clash with your document. Code blocks arrive as proportional text with their line breaks mangled. Lists sometimes convert and sometimes become paragraphs starting with a hyphen.
Cleaning that up on a long conversation takes longer than the conversation did.
Paste-as-plain-text avoids the styling mess and throws away all structure, so you lose the code blocks and headings that made the conversation worth keeping.
What proper conversion produces
A generated .docx should use Word's own semantics rather than imitating a web page:
Speaker turns as Heading 2, so the navigation pane gives you an outline
Code in a monospaced style with a light background, wrapped not clipped
Real Word lists, so numbering continues correctly if you edit
Real Word tables, editable rather than pasted as text
The conversation title as Heading 1 with the date beneath
Default body font, so it inherits your template rather than fighting it
Because those are native Word constructs, everything downstream works: the table of contents generates, styles apply, track changes behaves, and the document merges cleanly into a larger report.
Converting
Request your archive from ChatGPT under Settings, then Data Controls, then Export Data. Load conversations.json into ChatExports, select the conversations you want, and choose Word.
Each conversation downloads as its own .docx, named from its title. You can also convert your entire history in one pass if you are building an archive.
Conversion is JavaScript running in your browser. The export file is not uploaded, which matters given it contains everything you have discussed.
Common uses
Research write-ups. Convert the conversations behind a piece of analysis and attach them as an appendix, so reviewers can see the reasoning.
Client deliverables. Where an AI conversation informed a recommendation, a clean Word transcript is more credible than a screenshot and easier to redact than a PDF.
Legal and HR records. Word documents can be annotated with comments and tracked changes, which matters when several people need to mark up the same transcript.
Course material. Conversations that worked well as explanations convert into handouts with light editing.
Editing after conversion
Since the output uses real Word styles, restyle everything at once by modifying the style rather than reformatting by hand. Change Heading 2 to alter every speaker label; change the code style to change every code block.
If you are merging several conversations into one document, use Insert, then Object, then Text from File, which preserves styles better than copy and paste between documents.
Frequently asked questions
Will the Word document be editable?
Yes. It is a standard .docx using native Word styles, lists, and tables, so everything can be edited, restyled, and tracked normally.
Can I convert multiple conversations into one Word file?
You can convert each separately and combine them in Word using Insert, then Text from File, which preserves styling better than pasting.
Does it work with Google Docs?
Yes. Upload the .docx to Google Drive and open it with Google Docs. Headings, lists, and tables carry over, though code block shading may render slightly differently.
Are images included?
Images from your export are embedded where the conversation contained them, provided they were included in the archive OpenAI produced.