Apple has published a support page listing which Apple Intelligence tools now add an invisible SynthID watermark to your images. The page also explains how anyone can check a photo for that mark. The change starts with iOS 27.2, iPadOS 27.2, macOS 27.2 and visionOS 27.2, so it reaches well beyond the iPhone.
Which Apple Intelligence tools add a SynthID watermark
Apple first announced SynthID marks for generated images at WWDC in June. When we covered the iOS 27.2 release, the change was described only as Photos and Image Playground. Apple’s new support page, dated November 5, is more specific:
- Photos: Clean Up, Extend and Spatial Reframing
- Image generation: Image Playground and Image Wand
Apple says the mark can’t be seen and doesn’t change how an image looks. A SynthID detector can still read it.
One quibble with the page itself. Its opening line suggests that anything you create or edit with Apple Intelligence gets the mark. The list below it covers only some tools, so go by the list.
Older photos stay unmarked until you edit them
Anything you made or edited before the update carries no watermark. Apple’s support page puts the rule this way:
Images you create or edit before updating to iOS 27.2, iPadOS 27.2, or macOS 27.2 won’t have a SynthID watermark until you edit them with one of the tools listed.
That cuts both ways. A real photo from 2024 becomes a marked image once you remove a stranger from the background with Clean Up. If you plan to post an edited family photo, a detector may now flag it as AI-modified.
How to check a photo for a SynthID watermark
You don’t need an iPhone for the check. It runs in Google’s SynthID Detector, which opened to everyone in October.
- Open any web browser and go to synthid.com.
- Upload the image you want to check.
- Follow the on-screen steps to see the result.
Keep in mind that checking means uploading the photo to a Google website. If it’s a private image you’d rather not hand to Google, skip the check.
What a clean result doesn’t tell you
A result with no watermark does not prove a photo is real. The detector reads SynthID marks, the system used by Google and its partners. Images from AI tools outside that system can come back clean.
Small edits may slip through as well. In MacRumors’ own testing, Clean Up only seemed to add a mark when the edit was large enough. Small fixes came back unmarked. Images changed with Extend and Reframe did carry the watermark in the same tests.
Apple’s page lists Clean Up with no size condition, so treat MacRumors’ finding as an early test result, not a documented rule. It’s also unclear whether the mark survives a screenshot or heavy compression. As of November 7, we haven’t seen that confirmed either way.
SynthID is not the same as Apple Reference Image
The iPhone 18 Pro and Pro Max have a separate, opt-in feature called Apple Reference Image. It proves a photo came from the camera, while SynthID flags AI creation or edits. According to iThinkDiff, Reference Image isn’t available at launch in China, and its capture side isn’t offered in the EU.
If you care about how your edited photos will be read online, the practical step is simple. Check a test image after using Clean Up or Extend, and decide before you share.
Source: Apple Support, Apple Newsroom, MacRumors
