Apple’s iOS 27.2 adds invisible SynthID marks to Photos edits

By Tom Whitfield

Apple’s iOS 27.2, released last week, adds invisible SynthID watermarks to images changed with Clean Up, Extend, or Spatial Reframing in Apple Photos. Image Playground and Image Wand receive the same treatment, while iPadOS 27.2, macOS 27.2 and visionOS 27.2 extend the change beyond the iOS release.

Which Apple tools add the watermark?

Apple names five tools in the affected set, describing how “some Apple Intelligence tools” will add an invisible SynthID watermark to images they create or edit. The list comprises Clean Up, Extend and Spatial Reframing in Photos, plus Image Playground and Image Wand.

Image Playground creates entirely new images, which most people would recognize as AI output. The consequential case for working photographers is narrower and more slippery.

A genuine photograph in the library can also acquire the mark after an AI-powered edit. Apple notes that “the SynthID invisible watermark is also applied to genuine photos in your Apple Photos library” when users apply those editing tools.

Takeaway: The editing tool triggers the mark, not whether the underlying photo is real.

Can clients see the watermark?

No visible marker appears in the edited image. Nor will other people see one unless they explicitly check the file with a SynthID detector. That distinction matters in client handoffs: a delivered photograph can look untouched while carrying an invisible SynthID marker.

Apple recommends going to synthid.com, where users can upload an image and “see its result.” The release gives no accuracy rates, detector-coverage guarantees, or results for compressed, converted or subsequently edited files.

Apple also does not list export formats that preserve the mark or explain whether other retouching software removes it. Those omissions make any claim of tamper-proof evidence premature. Studios that need a provenance answer should check the final delivery file instead of relying on a client’s unaided inspection.

Takeaway: Clients see no marker, so provenance checking has to be deliberate.

What changes for paid photo work?

For paid assignments, the relevant issue is what a client can disclose, not whether a retouching tool has become fashionable. A conventional photograph edited with Clean Up, Extend or Spatial Reframing receives the same invisible treatment as an image created with Image Playground or Image Wand.

That difference calls for clearer delivery records rather than an automatic price adjustment. If a client asks whether a file was AI-modified, photographers can keep the untouched original, document the edit, and check the delivery copy with the detector Apple recommends.

None of those steps establishes ownership or changes a usage license. The update as described specifies no fee, royalty structure, licensing change, camera body, or lens requirement; it therefore supplies no basis on its own for repricing a job or replacing hardware.

Takeaway: Treat the tag as a disclosure signal, not a pricing or rights instrument.

What remains unanswered?

The release, described only as arriving “last week,” gives no calendar date, detection-accuracy figures, export-format list, or long-term support commitment. It identifies four releases carrying the feature—iOS 27.2, iPadOS 27.2, macOS 27.2 and visionOS 27.2—but leaves key verification questions open.

Apple does not explain whether original library files change, whether every export preserves the mark, or how a detector result should be interpreted in a disputed delivery. It also makes no claim that Apple blocks, removes, or penalizes tagged images.

On the evidence provided, this is an AI-labeling step, not a complete authentication system. The sensible business response is to know which tools trigger the tag, disclose their use when provenance matters, and verify the final file when a claim depends on it.

Takeaway: Apple has shipped a labeler, not a complete verification system.

Going forward, the practical question will be whether studios can verify and explain those marks in delivered files, rather than simply whether a hidden tag exists.