Skip to main content
← All posts
A 30s East Asian woman in a charcoal blazer reads a long Microsoft Copilot transcript on a dark laptop screen in a dim modern office, warm coral accents on the display, brass desk lamp casting rim light, editorial Monocle magazine style.
·7 min read

How to Read Microsoft Copilot (365) Chat History in 2026

Where Microsoft Copilot stores chat history across Teams, Word, and the standalone app, how to export cleanly, and how to read long AI answers back.

Microsoft Copilot sits inside Teams, Outlook, Word, and the standalone Copilot app. Chat history lives in different places depending on which door you walked through, and most workers do not realise it is scattered until they need to find an old answer. The problem is not that Microsoft hides anything, it is that Copilot history is split across surfaces and tenant policies. This guide covers where Copilot stores conversations across the Microsoft 365 surface, how to export them cleanly, and how to open the files in a reader built for AI output. Everything below assumes a work account on a tenant that has Copilot licensed, with notes where the consumer product differs.

Where Microsoft Copilot 365 stores your chats

Your Copilot history lives in two main stores, and the split matters more than it looks. Chats you start in the standalone Copilot app or at copilot.microsoft.com, signed in with a work account, are stored in Microsoft Substrate, the same mail-adjacent service that holds Outlook data. Chats inside Teams, Word, Excel, and PowerPoint are logged against those app sessions and surface differently. That means a Copilot thread inside a Word document never appears in your standalone Copilot history, and the opposite is also true. If you cannot find an old answer in the Copilot web rail, check the app you were working in at the time.

Admin-managed tenants keep Copilot history for the lifetime of the mailbox by default. You can see this yourself. Open Copilot in a browser, scroll the left rail, and the full list goes back as far as your admin policy allows. Compliance teams can see every message through Purview eDiscovery, so treat these transcripts as business records, not scratchpads. Retention is often tied to the broader mailbox policy, which means a 7-year legal hold on email also catches your Copilot chats.

The standalone consumer Copilot at copilot.microsoft.com, signed in with a personal Microsoft account, stores chats separately in your personal profile. If you flip between a work and personal account on the same browser, you are looking at two different histories, which is a common source of confusion when a conversation seems to vanish. Consumer Copilot keeps chats for 18 months by default, unless you delete them manually. Enterprise and consumer histories never merge, even if you use the same device, so the chat you remember from Tuesday may be in the other profile.

How to export a Copilot chat

The cleanest manual route is the share button on any Copilot conversation. Open the chat, click the share icon at the top right, then pick export. Copilot gives you a PDF or a Word document, which is fine for archive purposes but awful for reading long technical answers. Code blocks lose their monospace styling, tables mangle, and math breaks. If the chat is short and prose-only, PDF is acceptable, but anything with structure needs a different route.

For a readable export, copy the full conversation into a markdown file. Copilot already renders answers in markdown under the hood, so a direct copy preserves most of the structure without any post-processing. The flow is almost identical regardless of which Copilot surface you are on, and it works for Teams, Word, and the standalone app alike. Save the file somewhere you will remember, because the point of exporting is to find the answer again six months from now:

  • Scroll to the top of the thread so the first message is visible
  • Select everything from the first prompt to the last response
  • Paste into a plain text editor like VS Code or Sublime Text
  • Save with a .md extension and a descriptive filename
  • Open it in a markdown reader that renders code, tables, and math

You lose some styling, but the Mermaid diagrams, fenced code, and KaTeX math expressions survive intact. The result is a clean source file you own, not a locked PDF sitting in OneDrive. Prose keeps its paragraph structure, and headings stay as headings instead of flattening into bold text. If a code block loses its language hint, add it back manually, because the reader needs it for syntax colouring. A one-minute cleanup beats fighting PDF reflow every time.

Enterprise admins have a stronger option. Microsoft Graph exposes a Copilot interactions API that returns chat metadata and content as JSON, scoped to a user. If you own your tenant or have an admin willing to help, you can pull your entire history in one call and keep clean source files. This is the right path for anyone doing analytics on their own Copilot usage, or building a personal archive outside the Microsoft estate. The Graph response is JSON, so convert each chat to markdown before storing, otherwise the archive is unreadable without a script.

Open the exported file in a reader built for AI output

The hard part of Copilot history is not getting it out, it is reading it back. A long Copilot answer with code, a table, two Mermaid diagrams, and a KaTeX formula renders as a wall of broken text in most note apps. Word and PDF strip the structure, Notion stores the raw markdown but renders code blocks with no copy button and no inline math, and most dedicated code editors ignore the prose entirely. You end up switching tools three times to read one answer, which defeats the point of exporting.

Prism MD is built for exactly this. Drop the exported .md into the app, and the typography adjusts to AI-length answers: wider measure for prose, monospace tuned for code, inline KaTeX for math, and a diagram pane for Mermaid. The outline rail jumps you to any H2 inside the conversation, which matters when a single Copilot thread runs 8,000 words. See how the reader compares to Notion for AI conversations for a side-by-side on the same transcript. The same reader opens exports from other tools, so your Copilot archive sits next to Claude, ChatGPT, and Gemini history in one library. One tool for every AI you touch beats one tool per vendor.

If you are coming from a terminal-first workflow, the same read flow applies. The Codex CLI quick guide and the Claude Code archive guide cover file paths and export patterns for developer tools, and the output lands in the same markdown shape. One reader, one library, every AI tool you touch. That consolidation is where the real time saving sits, not in any single export step. The read experience is the thing that compounds, because every future export opens the same way.

Keep your Copilot history searchable

Export discipline matters more than any single tool. A chat is only useful later if you can find it, so add a short header to every export: date, topic in three words, the Copilot surface it came from. Prism MD indexes the full body plus any header you write, so a six-month search for a specific client answer takes two seconds instead of twenty minutes clicking through the Copilot rail. Treat the header as metadata, not prose, and keep it under five lines so it never gets in the way of the actual answer. Tagging is optional, but a single line like "tags: client-x, pricing, Q4" at the top makes faceted search trivial later.

Two habits pay off every time. First, export at the end of a session, not in a batch at the end of the week, because the conversation is still fresh in your head and the three-word header writes itself. Second, keep client work and internal exploration in separate folders, because mixing them triggers the kind of accidental share that compliance teams dread. A quarterly tidy of the archive is worth 30 minutes, and it keeps the search surface clean. If a chat was useful once, it is probably useful twice, which is the whole case for keeping an archive in the first place.

Frequently asked questions

How long does Microsoft Copilot keep chat history? By default, Copilot keeps chats until your mailbox is deleted or your admin applies a retention policy that trims it. Standalone consumer Copilot keeps chats for 18 months unless you delete them manually. Check Purview retention policies for the exact number inside your tenant. If your organisation has a legal hold on email, that same hold usually catches Copilot history, so expect long retention windows by default.

Can my employer read my Copilot work chats? Yes, through Microsoft Purview eDiscovery and standard audit log tools. Any Copilot interaction signed in with a work account is a business record and discoverable in legal proceedings. Keep that in mind for sensitive drafts, and use a personal account for anything that is not strictly work. Compliance teams rarely browse chats casually, but the capability exists and gets used during investigations.

Does Copilot history sync between the Teams panel and copilot.microsoft.com? Partially. Chats started in the standalone Copilot app sync to the web client under the same work account. Chats inside Teams, Word, or Excel stay in those apps and do not appear in the standalone history list. If you need a single searchable archive across surfaces, exporting to markdown and keeping the files in one folder is the only reliable approach.

What is the fastest export for a long technical answer? Select the whole conversation, copy, paste into a plain text file, save with a .md extension. Open it in a markdown reader that handles code, tables, and math. The whole flow takes under a minute per chat and preserves the structure that PDF export destroys. Build it into your end-of-session habit and the archive grows without any extra effort.

Read every Copilot export in one place

Free to start — no credit card.

Try Prism MD free →

Related reading

Ready to read your own AI documents?

Open ChatGPT, Claude, Gemini, or any markdown file in the reader built for the way models write.

  • ✓Renders code, math & Mermaid out of the box
  • ✓Works offline once you've opened a doc
  • ✓Free forever for personal reading