VisionOS 27 ships gaze triggered snapshots, not a Gemini Live style always on camera, and that design choice is the template the rest of the smart glasses category will have to answer.
Apple's Vision Pro doesn't watch the world on its own. It looks at it when you ask it to.
That single design choice is the most consequential thing Apple announced at WWDC 2026, even if the keynote mostly framed VisionOS 27 as a list of features. The new Visual Intelligence, the headline AI addition to the $3,500 headset, uses gaze tracking to take a snapshot of whatever the wearer is looking at, then hands that still image to a rebuilt Siri for context. It does not run an ambient video feed. It does not watch continuously. The user pulls the trigger, in the form of an eye fixation, and only then does the camera engage.
Samsung and Google, by comparison, ship a headworn product that does the opposite. The Samsung Galaxy XR, the mixed-reality headset Samsung and Google launched last fall for $1,799, runs Gemini Live: a continuous, camera-aware assistant that sees what the wearer sees and answers in real time, including the Circle to Search in 3D space feature CNET has been testing on the device. Apple's choice to ship a gaze-triggered snapshot instead is a deliberate counter-pattern, and the rest of the smart-glasses category now has a concrete model to argue with.
That pattern is also the preview. Per CNET's WWDC coverage, CNET's Scott Stein characterizes Apple's expected smart glasses — which he reports are likely arriving next year — as the likely inheritor of this camera behavior. Visual Intelligence on the headset reads like an early test of the snapshot model those glasses will inherit. If the snapshot model survives the transition to a smaller, lighter, more socially wearable device, it will set a default that competitors will have to either match or rebut. If Apple flips to continuous capture later, the headset is a hedge, not a template.
The rest of VisionOS 27 is best read as scaffolding around that core decision. According to CNET's WWDC live blog, Apple announced that personal panoramic photos can be wrapped into 3D immersive backgrounds, that 3D objects from Mac apps can be previewed inside the headset, and that a new app-extension workflow lets a Mac and a Vision Pro share a task. None of those additions change the device's relationship to the wearer's camera. They are texture.
The more substantive change is Siri itself. The new "Siri AI" is a single cross-device app with on-screen awareness, conversation memory, and the ability to identify people, places, and objects in photos. CNET reports that the system is powered in part by Google Gemini models, an unusual admission for Apple, and quotes Mike Rockwell, the executive who led the Siri AI effort, describing the new behavior on stage. The same app surfaces across iPhone's Dynamic Island, Mac's Spotlight, and the headset. The new Siri is the brain; Visual Intelligence is one of its inputs; the gaze-snap is the gate.
A few caveats matter. Every direct quote from Apple used here comes via CNET, not from an Apple press release: Apple's own newsroom page for the event returned a 404 when checked, so the language is "per Apple at WWDC" as reported by CNET, not verbatim from an Apple document. The public beta timing, framed by Apple at the keynote as "next month," is the standard WWDC pattern, so July is the working assumption rather than a confirmed date. The $250 million Siri/Apple Intelligence settlement that CNET references in its WWDC write-up as having been paid "just last month" is reported, not adjudicated in the materials used here, and any precise settlement date is out of reach. Scott Stein, CNET's Editor at Large who has covered VR and AR for two decades, calls VisionOS 27 "mostly incremental except for Siri" and frames Visual Intelligence as groundwork rather than destination. That is one editor's read, not an industry consensus, but it is the read of someone who has been on this beat longer than most of the category has existed.
The honest limitation travels with the story. Stein's point is that 3D backgrounds, Mac previews, and the cross-device app extensions are unlikely, on their own, to convert someone who has not already decided the $3,500 headset is for them. They are reasons for an existing owner to update, not reasons for a curious outsider to buy. Visual Intelligence is a different kind of feature because it ships with a question attached: should a headworn camera that sees what the wearer sees fire continuously, or only on intent? Apple has answered that question for the headset. The smart-glasses category will spend the next eighteen months arguing about whether the answer scales.
The leadership transition sharpens the timing. WWDC 2026 was Tim Cook's last as Apple CEO; CNET reports that John Ternus takes over in September 2026. Whoever signs off on the smart-glasses camera behavior, the spec is being written now, on a device that still targets a niche enthusiast audience rather than the mainstream market. That makes VisionOS 27 less of a product story and more of a public draft of a category-defining choice.