Key Takeaways
- Almost everyone asks AI about their screen the same way: screenshot it, upload it, then explain it. That is six steps for one question.
- You can skip all six. Hold Control and Option, move your mouse over the part you are asking about, and speak.
- Your mouse leaves a blue mark on the screen. That mark is how the AI knows what you mean, so the word “this” suddenly works.
- It sits on top of any app. Gmail, a bank statement, a settings page, a photo, a spreadsheet, a full screen video.
- Your screen is only looked at while you are holding the keys and asking. Nothing runs in the background.
The six steps you do without thinking
An email lands. Something about the tone feels off and you are not sure how to answer it. So you do what everyone does. You press the screenshot keys, drag a box around the message, open ChatGPT, drag the picture in, and type a sentence explaining what you just sent.
That is six steps to ask one question. You did it because there was no other way to show the AI what you were already looking at.
Millions of people repeat this every day. Screenshots of emails. Screenshots of error messages, bank statements, homework, and settings pages where a button should be and is not. The screenshot was never the point. It was just the only envelope we had.
What gets lost the moment you crop
Cropping is guessing. You have to choose a box before you know what the AI needs to see. Cut too tight and the important part is outside the frame. Cut too wide and you end up explaining which bit you meant anyway.
Then comes the typing. “This is an email from my landlord about my deposit, and I want to sound firm but not rude.” You just described, in words, something that was sitting right in front of you.
That is the real cost. Not the thirty seconds. The fact that you had to translate what you were looking at into a paragraph before anyone could help you.
Point at it instead of describing it
Here is the whole idea. Stop describing what you mean and point at it, the way you would if a friend were sitting next to you.
In VoiceOS you hold two keys, Control and Option. Your cursor turns into a crosshair and a small label appears beside it that reads “Point to show VoiceOS”. Now move your mouse over the thing you are asking about, and talk.
As you move, a blue trail follows your cursor like a highlighter. You are not clicking. You are not dragging anything. You are just sweeping the mouse over the part that matters while you speak.
Let go of the keys and VoiceOS hands over one picture of your screen with your mark on it. That is how it knows what “this” refers to. The vague word stops being vague.
Can you find his LinkedIn
What people actually say out loud
Look at how short these are. You never set the scene, because the scene is already on the screen.
“How should I reply to this?”
Circle the message. It reads the thread around it too.
“Does this sound too harsh?”
Sweep over your own draft before you hit send.
“What does this error mean, in normal words?”
Works on any warning box, in any app.
“Where do I turn this off?”
Point at the settings page you are lost in.
Five moments this replaces a screenshot
The reply you keep rewriting. A message from a coworker, a landlord, another parent from school. Sweep over it and ask how to answer. Or point at your own draft and ask whether the tone lands right.
The envelope you did not want to open. A bill, an insurance form, a letter from the tax office. Mark the paragraph that makes no sense and ask what it actually means for you.
The button that is not there. A guide says click the thing, and the thing is nowhere on the page. Point at the page and ask where it is hiding.
Homework at the kitchen table. A chart, a formula, a page of a textbook on screen. Point at the one line that broke your brain instead of photographing the whole page.
Two tabs and a decision. Open both options side by side, sweep across them, and ask which one is the better deal.
No, it is not watching your screen all day
This is the first thing people ask, and it is the right thing to ask.
Your screen is looked at while you are holding the keys and asking a question. Let go and that stops. There is no recording running in the background and no timeline of your day being saved somewhere.
The feature also arrives switched off. You turn it on yourself in settings, under “Ask about your screen”, and that is the moment your Mac asks for screen permission. Nothing happens before you say yes.
One detail people enjoy: the blue trail your cursor leaves does not show up in screen recordings or screen shares. You can use it in the middle of a call and nobody on the other side sees it.
Try it in the next two minutes
Open the thing you were about to screenshot. An email, a form, a page that confuses you. Anything.
Hold Control and Option. Wait for the crosshair. Move your mouse over the part you care about and ask your question in the plainest words you have.
Let go, and read the answer. You just skipped the crop, the upload, and the paragraph of explanation.
Frequently Asked Questions (FAQ)
Do I still need screenshots to ask AI about my screen?
No. Hold Control and Option, move your mouse over the thing you are asking about, and speak. VoiceOS receives one image of that moment with your mark on it. There is nothing to crop, save, or upload.
How is this different from sending a screenshot to ChatGPT?
Two things change. You no longer choose the crop before you know what matters, and the word “this” starts working. With a screenshot you pick a box and then describe what you are looking at. When you point, the mark is already on the picture, so a three word question is enough.
Do I have to draw a neat circle?
No. Any movement over the thing is enough. A loop around it, a line under a sentence, a single dot on a button. There is no drawing skill involved and nothing is graded.
Does it work in every app?
Yes. It is not a feature inside one app, it draws on top of whatever is already on your screen. That covers browsers, email, PDFs, spreadsheets, chat apps, and full screen apps, with no per app setup.
Is my screen being recorded?
No. Your screen is only captured while you are holding the keys and asking. There is no always on recording. The feature also ships off by default, so it does nothing until you enable “Ask about your screen” in settings and grant permission.
What if I use two monitors?
The display you move across is the one that gets marked, and it can mark up to three screens in one question. So you can point at a chart on one monitor and a table on the other and ask about both together.
Does this work on Windows?
VoiceOS itself runs on Mac and Windows. Point and ask is built and tested on Mac today. On Windows, start with voice typing and Agent Mode, which work there now.
Next time you reach for the screenshot key, point instead
VoiceOS runs on Mac and Windows. Ask about anything on your screen, in any app.
Download VoiceOS