Visual Intelligence on Mac: What It Does, and What Still Needs a Screenshot Tool

The short version. Visual Intelligence is a question-and-answer layer over your screen. You select a region and ask Siri about it, or get an action button when Siri recognises a URL, a date, a phone number, or an address. It is good at understanding. It produces no image, so it cannot annotate, redact, scroll-capture, or give you a file to send. And it needs an Apple Silicon Mac on macOS 27, which leaves every Intel Mac out.

What Visual Intelligence actually is

macOS 27 Golden Gate was released on 14 September 2026. It put Siri AI behind a new screen-selection tool: drag a box over part of your screen, a window, or an image, and you get a small panel with Ask Siri, Image Search, and any action Siri thinks applies to what it found.

It is the same idea Apple brought to the iPhone camera: point at a thing, ask about the thing. On the Mac the “thing” is whatever is already on your display, which is usually more interesting than what is in front of your camera. A stack trace, a chart in a deck, a paragraph in a language you do not read, a form field you cannot parse.

How to trigger it

• ⌘⇧6 gives you the selection crosshair directly. Drag, release, and the panel appears. This is the fast path for everyday use.
• ⌘⇧5 opens the familiar screenshot toolbar. Hit the Siri icon, then Ask Siri. Use this when you already wanted a window or full-screen capture rather than a hand-drawn box.

There is no file at the end of either path. That is the whole design: you asked a question, you got an answer.

Where it is genuinely good

Reading unfamiliar interfaces. Asking “what does this error mean” about a dialog you have never seen is the best use of the feature, and it is faster than retyping the error into a chat window. The same goes for a chart with no caption, or a settings screen in another language.

It is also good at turning pixels into structured details. A date on a poster becomes an Add to Calendar button. A phone number becomes a contact. A URL becomes a link you can open without retyping it character by character, which sounds trivial until you have mistyped a tracking number twice.

Where it stops

The gaps are not bugs. They follow from the feature producing answers instead of assets.

• No output file. There is nothing to attach to a ticket, paste into a doc, or keep as evidence. If your job is “show the designer exactly which button is wrong,” an answer is the wrong shape.
• No annotation. No arrows, no numbered callouts, no boxes. The box you drew disappears once you release the mouse.
• No redaction. It cannot blur a client email or a key before the image leaves your machine, because there is no image leaving.
• No scrolling capture. It sees the viewport. A long thread, an invoice list, or a full-page dashboard is beyond the selection.
• No screenshot history. Ask about the same region tomorrow and you start over. Nothing is indexed, nothing is searchable.

There is one more caveat worth taking seriously: the answers are not always right. Asked for related wallpapers of a specific car, Apple’s own reviewer got photographs of a different model. Treat a confident answer about a food item, a plant, or a part number the way you would treat any language model output, and check the values that matter against the original.

The Intel problem nobody mentions

macOS 27 does not run on Intel Macs at all. Apple’s compatibility list for Golden Gate is Apple Silicon only, and macOS 26 Tahoe is the last release with Intel support. Siri AI adds a second gate on top: it needs an Apple Intelligence capable Mac, in practice M1 or later, plus a supported language and region.

So for a large group of people still working on a 2018 to 2020 Mac, Visual Intelligence is not a feature they are choosing not to use. It does not exist on their machine, and will not arrive. That is a good reason to look at what third-party tools still run on macOS 26 and older rather than reading nine articles about a feature you cannot install.

So do you still need a screenshot tool?

If your only goal is “tell me what this says,” no. Visual Intelligence covers that, and the built-in shortcuts cover a plain capture. Use them.

You need something else when the job ends in a file that has to be marked up, cleaned, or handed to another person. That is the half of the workflow Apple did not touch. A capture tool like iShot stays in that half: it captures, lets you draw on it, blurs what should not be shared, stitches a long page into one image, and puts the result on your clipboard. Text extraction runs on Apple’s Vision framework locally, so the reading half is the same technology Visual Intelligence uses, without the requirement to send anything anywhere.

iShot runs on macOS 12 and later, as one Universal build that covers both Apple Silicon and Intel. That last part matters more this year than it did last year. If you are on an Intel Mac, you are not getting macOS 27, but you can still run a current screenshot tool.

Frequently asked questions

Does Visual Intelligence replace a screenshot app?

For questions, mostly yes. For output, no. It explains what is on your screen but does not produce an image you can annotate, redact, save, or send, and it has no capture history.

Which Macs support Visual Intelligence?

Apple Silicon Macs running macOS 27 Golden Gate with Apple Intelligence enabled. Siri AI was in beta at release and needs a supported language and region. Intel Macs cannot install macOS 27, so they never get the feature.

Can I still take a screenshot on macOS 27 the old way?

Yes. ⌘⇧3, ⌘⇧4, and ⌘⇧5 work exactly as before, and Preview still opens images for markup. Visual Intelligence was added alongside those tools, not in place of them.

Is my screen content uploaded when I use Visual Intelligence?

Apple states that Apple Intelligence can process requests on device or use Private Cloud Compute for more complex ones, and Image Search goes to Google. Check your Siri settings before selecting anything containing credentials or other people’s data.