Getting Started With Copilot Vision: A Real Walkthrough
Subscription required
Copilot Vision requires a personal Microsoft 365 subscription (Personal, Family, or Premium). It's not part of the free Copilot experience, so check your subscription before you build a habit around it.
Talking to Copilot is one thing. Showing it what's actually in front of you is a different kind of useful, and it's easy to miss entirely because it's not the feature Copilot leads with. Copilot Vision lets Copilot look at your screen, or your phone's camera, while you're in a Voice conversation with it, so instead of describing a problem in words, you can just point at it.
If you haven't used Copilot's Voice mode before, the Complete Beginner's Guide to Copilot covers the basics of starting a voice conversation, which this article builds directly on.
Setting it up
- 1
Confirm your subscription
Vision is gated behind a personal Microsoft 365 subscription. If you don't have one of those plans, Vision won't be available to you, so this is worth ruling out first if you don't see the option.
- 2
Start a Voice conversation
Open Copilot and start a Voice session the way you normally would. Vision isn't a separate mode you switch into from scratch, it's an extension of a Voice conversation you're already having.
- 3
Turn on screen or camera sharing
Inside the active Voice session, there's an option to share your screen. On desktop you select Share screen and pick the app or screen you want Copilot to see; on mobile you switch to the camera instead. Choose whichever matches what you want Copilot to see.
- 4
Talk while pointing, don't just describe
Once sharing is on, the whole value of Vision is that you can say "see this error message here" or point the camera at something, instead of typing out a description of what you're looking at. Talk to it the way you'd talk to someone standing next to you looking at the same thing.
- 5
Turn sharing off when you're done
Sharing stays active only for that Voice session, and Microsoft says each Vision conversation is independent, with no long-term recall of what was shown in an earlier session. Ending the conversation or stopping the share stops Copilot from seeing anything further.
Be mindful of what's on your screen
Copilot sees everything visible in the app, window, or screen you choose to share, not just the specific part you're asking about. Close anything sensitive, another tab with financial details, a private message thread, before you start sharing, the same way you would before a screen share on a video call.
Three real situations where it's worth using
Troubleshooting software you don't know well. Say a spreadsheet a colleague sent you has a formula throwing an error you don't recognize, and you don't have the vocabulary to describe the problem in a text prompt. Share your screen, point at the cell, and ask "why is this showing an error, and what's it actually trying to calculate?" Copilot can see the formula and the surrounding cells directly, instead of you retyping the formula by hand and hoping you copied it correctly.
Following a recipe with your hands full. Point your phone's camera at a recipe card or a cookbook page, then keep talking to Copilot hands-free while you're mid-recipe: "how many minutes did it say for the second step" or "what's a substitute for the ingredient I don't have." This is a case where Vision genuinely beats reading a screen yourself, because your hands are covered in flour and a voice conversation doesn't require touching anything.
Reviewing a visual design before it ships. A marketer looking at a draft social graphic or a slide layout can share their screen and ask Copilot to react to it directly: "does the text in the top corner look like it's competing with the logo?" or "is there enough contrast between this text and the background?" Copilot responding to what it's actually seeing, rather than a written description of the layout, catches things a text-only prompt would miss entirely.
Tip
Vision is strongest as a second pair of eyes on something visual and ambiguous, not as a way to avoid typing. If a question is easy to describe in one sentence, typing it is faster than setting up screen sharing.
A worked example: talking Copilot through a spreadsheet error
Take the troubleshooting situation further, since it's the clearest illustration of what Vision adds. Say a formula is returning an error on some rows and not others, and there's nothing obviously wrong with it at a glance. The exchange below is illustrative, meant to show the shape of the conversation, not a captured transcript.
A Vision conversation about a spreadsheet formula error, illustrated
Why this genuinely needs Vision, not just a good text prompt
Notice what never had to happen here: you didn't type out the formula, the column layout, or the exact contents of two cells, all details that are easy to get slightly wrong when copying by hand, and a wrong detail in a text prompt produces a wrong answer with just as much confidence as a right one. A trailing space in particular is close to invisible to a person scanning a cell quickly; it's the kind of difference Copilot can catch by actually looking at the data, but that a text description would almost never think to mention, because you'd have to already suspect it to describe it. That's the real gap between Vision and text-only Copilot: text-only Copilot is only ever as good as your description of the problem, while Vision can notice the thing you didn't know was worth mentioning.
Where it falls short
Vision reads what's on screen or in frame at a given moment; it doesn't retain a deep understanding of a document the way uploading a file to a chat would. Microsoft notes real limits at this stage: it can't read video or animated content, and switching between windows too quickly while asking a question can cause it to answer based on the wrong screen. It's also not a substitute for actually opening and inspecting a file yourself when precision matters, if a number in a spreadsheet needs to be exactly right, verify it directly rather than trusting a spoken read-back from a screen share. And because it depends on a live Voice session, it's not built for a quick, silent, at-your-desk question; if you'd rather type than talk, a normal chat with an attached screenshot is often the faster path to the same answer.
Vision requires a personal Microsoft 365 subscription and an active Voice conversation
Close anything sensitive before sharing your screen; Copilot sees everything in what you share
Best for visual, ambiguous, hands-full situations, not as a replacement for typing a quick question
Verify anything precise, like an exact number in a spreadsheet, yourself rather than trusting a spoken read-back
Official sources
Checked on September 21, 2026. Features, plans and names change often, so the vendor's own pages are the final word.