Copilot Vision lets you share your desktop screen or your phone camera with Microsoft Copilot and ask about what it sees. To use it, start a voice chat with Copilot, then select the screen share arrow on desktop or the camera icon on mobile. It needs a Microsoft Copilot licence on a work account, or Microsoft 365 Personal, Family or Premium on a personal account.
Vision only works by voice. That surprises most people, who go looking for an upload button in the text chat. You talk, Copilot looks at your screen or camera, and it answers out loud. Here's how to set it up and get good answers, from Microsoft's own support pages, checked as of October 2026.
What you need for Copilot Vision
| Requirement | Work or school account | Personal account |
|---|---|---|
| Licence or subscription | A Microsoft Copilot licence | Microsoft 365 Personal, Family or Premium |
| Device to share a screen | Windows or the web | Windows or the web |
| Device to share a camera | Microsoft Copilot mobile app | Microsoft Copilot mobile app |
| Microphone | Required | Required |
| Camera permission | On mobile | On mobile |
Microsoft says vision availability may vary while features roll out, and it works in all the languages Microsoft Copilot supports.
How to start Copilot Vision
- Start a voice chat with Copilot. On Windows you can press the Copilot key or Win+C.
- On desktop, select the screen share arrow. On your phone, tap the camera icon.
- The first time, Copilot may ask you to acknowledge a privacy notice.
- Ask your question out loud, the way you'd ask a colleague sitting next to you.
- To stop, select Stop sharing on desktop or tap the camera icon again on mobile. Your voice chat carries on.
Prompts that work well with vision
Microsoft suggests starting with questions about what's on screen. These three are adapted from its own starter list:
For a report or dashboard
What are the key trends in this dashboard, and how do they compare with the report I reviewed last month?
For design feedback
Look at this banner and the feedback I received. What changes should I make first?
When you're stuck in software
I'm stuck on this screen. Walk me through the next steps to complete this task.
On your phone, Microsoft's example is pointing the camera at a device error and asking how to fix it. I'd add product labels, a printed invoice or a whiteboard after a planning session.
How vision works behind the scenes
- You share visual input during a voice conversation: your desktop screen on Windows or the web, or your phone camera.
- Copilot converts it into data it can analyse, including text, images, charts and on-screen interfaces.
- On a work account, it combines what it sees with your Microsoft 365 data, such as documents and emails.
- It answers by voice in real time, so you can ask a follow-up straight away in the same conversation.
What Copilot Vision can't do
This is the part to read before you rely on it. Microsoft lists these limits:
- It can't take actions. Copilot might highlight parts of the screen, but it won't click, type or scroll for you.
- It has no long-term recall. Each vision conversation is separate, and it doesn't reuse screens or camera input from earlier chats.
- It only works inside voice chats for now.
- Daily use depends on available capacity and counts toward your daily voice usage. The Copilot app warns you as you get close.
- Switching windows too quickly can make Copilot answer about the wrong screen.
What Microsoft does with what you share
Here's the insight most guides skip. Microsoft says the screen or camera content is processed as a series of images, not a continuous recording. Audio and video data are stored temporarily so you can give feedback, then deleted after 48 hours. Text transcripts of the voice conversation are stored the same way as your normal Copilot text chats.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
On a work account, Copilot also grounds its answers in your Microsoft 365 data, such as documents, emails, meetings and earlier discussions. That's powerful when you ask "what feedback did I get on this?", but it's also a reason to close anything confidential before you share your whole screen. Microsoft adds that vision doesn't infer sensitive personal attributes such as race or emotion, and that thumbs up or down feedback isn't used to train the foundation models behind Copilot.
Tips for better answers
- Keep the thing you're asking about on screen until Copilot has answered.
- Check your microphone is on before you start, because vision works through voice.
- Ask one question at a time and name what you're pointing at.
- Share a single window rather than a desktop full of open apps.
Should you use it?
If you've ever wished you could turn to someone and ask "what am I looking at?", vision is the closest Copilot gets. It's best for explaining dense reports, talking through feedback, and getting unstuck in software you don't use often. It isn't an agent, so don't expect it to do the clicking. If you're not sure your plan includes it, start with what you get with free Copilot, read my guide to using Copilot day to day, and if you're on a work account, see how Microsoft 365 Copilot uses your work data.
My verdict on Copilot Vision
A useful second pair of eyes, with one big caveat: it can see everything you share and, on a work account, connect it to your files and email. Share one window, ask clear questions, and you'll get far more out of it than people who share their whole screen and hope.
Once you've got an answer worth keeping, read how to use Copilot Pages to see how to save it.
Frequently asked questions
How do I turn on Copilot Vision?
Start a voice chat with Copilot, then select the screen share arrow on desktop or the camera icon on mobile.
Is Copilot Vision free?
Microsoft says vision needs a Microsoft Copilot licence on a work account, or Microsoft 365 Personal, Family or Premium on a personal account.
Can Copilot Vision click or type for me?
No. Microsoft says vision can highlight parts of the screen but won't click, type or scroll.
Does Copilot Vision record my screen?
Microsoft says shared content is processed as a series of images, and audio and video data are deleted after 48 hours.
Does Copilot Vision remember earlier sessions?
No. Each vision conversation is independent and doesn't reuse screens or camera input from previous chats.
Can I use Copilot Vision in text chat?
No. Microsoft says vision is only available during a voice conversation with Copilot right now.
