iOS 27.2 adds hidden SynthID watermarks to iPhone AI images

What has changed

Apple's iOS 27.2 update now puts a hidden SynthID watermark on images created with Image Playground and on some photos edited with Apple Intelligence, according to MacRumors. Apple announced plans to add SynthID support in June, and the feature is now live in iOS 27.2 ^[1].

SynthID is a tool created by Google to identify AI-generated content. Instead of adding information to a file's metadata, it marks images and videos at the pixel level. Google says it does not affect image quality and is meant to survive cropping, filters and compression. MacRumors notes that SynthID is used by Google, Apple, OpenAI and others ^[1].

Which images are marked

MacRumors tested the Photos app tools and reported mixed results. Images altered with Extend and Reframe carried SynthID watermarks when checked. The Clean Up tool was less consistent: the watermark seemed to depend on the size of the edit. Small edits did not receive one, while larger Clean Up edits did. Images created in Image Playground all carry the watermark ^[1].

That distinction matters. A lightly retouched photo may still be treated as a normal picture, while a heavily changed one is flagged as AI-altered. Anyone relying on the watermark as proof that a picture is untouched should be aware of that gap, which is based on the publication's own testing rather than a statement from Apple.

No built-in way to check

There is currently no option on the iPhone for detecting SynthID watermarks. To check an image, you need to use the SynthID Detector website ^[1]. In practice, that means the mark is aimed at platforms, journalists and fact-checkers who run images through a detector, rather than at everyday iPhone owners scrolling through their photos.

Why it matters for businesses

Many companies now use AI image tools for marketing mock-ups, product shots and social posts. A hidden watermark that survives common edits makes it easier for others to establish where an image came from. That can help with trust: a business that publishes AI-assisted pictures can be identified as having done so, and a business that receives images from suppliers or customers has one more way to test them.

There are limits. The marking only applies to content made with tools that use SynthID, and the sources do not suggest it can identify images from every AI generator. Readers who commented on the MacRumors piece raised the same point about other tools, but that is reader opinion rather than reporting.

What to do now

If your team uses Image Playground or Apple Intelligence editing, assume the output may be identifiable as AI-generated, and update any internal guidance on disclosure accordingly. If you handle image authenticity, for example in insurance, recruitment or media, it is worth knowing that a free detector exists and that results can vary with the size of the edit.

The change arrives with iOS 27.2, which MacRumors describes as the first major update to iOS 27 ^[1]. Updating is the way to receive the watermarking. Apple has not, in the material we read, said whether it will add a detection tool to the iPhone itself.

For now, the practical takeaway is simple: Apple's AI images are no longer anonymous, even though only an outside website can tell.

Source: MacRumors — iPhone AI Images and Edits Now Have a Hidden Watermark