Copilot: View Memories and Review What It Remembers
Find the memory controls for your Copilot account, review saved details, and correct suggestions before treating them as project facts.
Updated 5 min read
To view Copilot memories, open the memory management settings for the Copilot product and account you actually use. Microsoft’s personal Copilot app, Copilot with a work or school account, and GitHub Copilot have different controls, so there is no single route that fits all three.
Reviewing a saved detail helps you catch outdated preferences and guesses that sound more certain than they are. Tools like fenbs keep confirmed project facts and rules outside one assistant, so you can review the work separately from an app’s personal memory.
Which Copilot are you checking?
Start by checking the signed-in account and app name. A personal Microsoft account and an employer’s work account may appear in the same browser. Looking at the wrong profile can make a perfectly ordinary difference look like missing memory.
| Your product | Start here | Avoid this assumption |
|---|---|---|
| Microsoft Copilot for individuals | Personal account settings and Personalization. | A work account has the same menu. |
| Copilot Chat with a Microsoft 365 account | Chat settings and Personalization. | Saved memories contain every personalization source. |
| GitHub Copilot | GitHub’s Copilot Memory documentation and account settings. | Microsoft Copilot app controls also manage coding memory. |
If the product names still blur together, our Copilot and ChatGPT comparison gives broader orientation. This guide stays focused on inspecting what is remembered rather than choosing an assistant.
How do you open saved memories in personal Copilot?
- Sign in to Microsoft Copilot with the personal Microsoft account you want to review.
- Open Settings, then Personalization.
- Select Manage next to Saved memories and read the entries shown there.
- Delete an inaccurate or unwanted entry using its delete control. Use Delete all memories only if clearing the entire list is your intention.
These are the steps in Microsoft’s current personal privacy controls (opens in a new tab). Its older app instructions (opens in a new tab) instead describe Memory controls under the profile menu and asking what Copilot knows about you. Match the instructions to the screen you have.
Before changing anything, decide whether the entry is wrong, merely old, or still useful for a limited purpose. For example, “prefers short answers” might be helpful personally but unsuitable for a detailed research assignment. A preference should not silently become an instruction for every kind of work.
Where are memories in work or school Copilot Chat?
Microsoft’s Copilot Memory management guide (opens in a new tab) describes Settings and more, Chat settings, Personalization, then Manage saved memories. You can remove individual entries or clear the list. Check your account before applying that route to a personal app.
The same guide describes asking Copilot to remember a detail and receiving a Memory updated notice. If Copilot suggests saving something, read the proposed meaning before accepting it. A casual statement about today’s assignment may be a poor standing preference.
Say you are preparing a hypothetical client presentation and ask for very brief slide notes. That does not necessarily mean you always want brief analysis. When reviewing a suggestion, narrow it to the situation you intended or decline it if it adds no lasting value.
For an approved project instruction, fenbs Decisions and rules gives you a separate place to record what a person actually chose and where it applies.
How should you check and correct a suggestion?
Treat a suggested memory as a draft summary. Look for the difference between your words and the generalization being saved. “Use Spanish for this email” and “always respond in Spanish” have very different effects on the next task.
- Accuracy: does the entry say something you actually told Copilot?
- Scope: is it about you, one project, or a temporary assignment?
- Authority: was it an approved decision or just an option you discussed?
- Usefulness: will keeping it improve a future answer?
- Sensitivity: should this information be in an assistant’s saved context at all?
Remove an entry you do not want retained. If you want a replacement, state the corrected detail plainly and check the resulting saved entry rather than relying on a conversational acknowledgment. Keep the test harmless; you do not need to repeat confidential information to check whether a control works.
A useful review record might say: “Removed the old preference for slide summaries; detailed written analysis remains the default for this project.” Save only what the next person needs to know. Recording the removed sensitive detail again would defeat the purpose of the cleanup.
Keep an approved rule separate from a suggestion
Record the project instruction a person approved in fenbs Decisions and rules, with its scope and reason. Connected assistants read rules first when they retrieve project context. That gives the next session an explicit work record to check against a remembered preference.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
Does the memory list explain every personalized answer?
Not necessarily. Microsoft’s personalization explanation (opens in a new tab) distinguishes saved memories from chat-history inferences, custom instructions, and work insights where available. A saved-memory review is useful, but it is not a complete audit of every source an answer could use.
Check what is in the current conversation as well. If the answer repeats a detail you just supplied, that alone tells you little about saved memory. A fresh, ordinary conversation can help distinguish the current thread from a longer-lived preference, although a model’s answer still is not proof of backend deletion.
If your issue concerns written directions rather than remembered details, see Copilot instruction examples. Keep the two questions separate while troubleshooting so you change the control responsible for the behavior.
What if you meant GitHub Copilot memories?
GitHub documents Copilot Memory (opens in a new tab) as repository facts and personal coding preferences. Users can review their preferences; repository owners can review repository facts. Use GitHub’s linked management guides for the controls that match your role.
Do not try to remove a coding preference through the personal Microsoft Copilot app. For orientation on the developer product, read our GitHub Copilot app guide, then return to the memory settings in the correct service.
How to do this with fenbs
- Create a free account, then add the project whose facts you want to keep consistent.
- Write confirmed project facts as AI context notes. Record approved instructions in Decisions and rules, including who decided and why.
- Use Integrations to connect a supported assistant over MCP, a standard way for an AI app to use an outside tool. GitHub Copilot is supported; the consumer Microsoft Copilot app is not.
- Ask the connected assistant to read the project context before work and save progress before stopping. Review changes in History when you need to check who updated the record.
fenbs does not inspect, edit, or import Copilot memories. It is not the right fit if you only want to change a personal preference inside Microsoft Copilot. It becomes useful when confirmed work context must also reach another supported assistant; our AI memory hub covers that wider problem.