iOS 27.2 Photos SynthID watermark: How it actually works

Update to iOS 27.2 and the edited photos in a camera roll look exactly the same as before. That's intentional. Apple's latest release, which rolled out last week, quietly embeds an invisible marker called SynthID into certain AI-generated and AI-edited images, so a photo can carry a trace of its AI origins without any visible change to the picture itself (Apple Support, last week; 9to5Mac, this week).

The iOS 27.2 Photos SynthID watermark doesn't touch everything in a library. It applies to Image Playground creations, the Image Wand tool, and three specific editing tools inside Photos, Clean Up, Extend, and Spatial Reframing, across iOS, iPadOS, macOS, and visionOS 27.2 (9to5Mac, this week). Because the mark lives in pixel data rather than metadata, Apple built it to survive cropping, filters, and screenshots in ways a metadata tag typically can't (Cult of Mac, last week).

Here's which edits receive the label, how to check one, and what a result actually tells you. Anyone who already updated, or is deciding whether to, should know exactly what changed in their Photos app and what it doesn't prove.

How does the iOS 27.2 Photos SynthID watermark work?

SynthID is Google's invisible, pixel-embedded marker, created to help identify AI-generated content as that kind of content gets harder to spot by eye (Cult of Mac, last week). Think of it less like a stamp sitting on top of a photo and more like a signature woven into the image data itself: invisible to the eye, readable only by the right detector.

Apple's iOS 27.2 release notes confirm the adoption, listing SynthID support in Photos and Image Playground specifically to help identify whether an image was generated or modified with generative tools (Apple Support, last week). Apple flagged this plan at its September event, saying SynthID support would land in a software update later in the year, and iOS 27.2 is that update (iThinkDiff, last week). The company signaled the integration even earlier, at WWDC26 in June, alongside a redesigned Image Playground (Cult of Mac, last week).

The practical limitation is that Photos cannot scan for the mark itself. Apple adopted Google's existing watermarking standard rather than building its own scanner into the app, so confirming whether a specific photo carries the label means stepping outside Photos entirely and using a separate tool, covered further down.

Which Apple tools add a SynthID watermark on Apple Photos edits?

Coverage splits across three tiers, and which tier an edit lands in matters more than just knowing the feature exists.

Always watermarked: Everything generated from scratch in Image Playground carries the mark, along with images made using the Image Wand tool (9to5Mac, this week; Cult of Mac, last week).

Consistently watermarked, per testing: Extend and Spatial Reframing inside Photos reliably trigger the mark, since both generate substantial new image content rather than making small tweaks (iThinkDiff, last week; 9to5Mac, this week).

Conditional, depending on edit size: Clean Up is the inconsistent one. Testers found that small object removals often go unmarked, while larger edits do trigger the watermark, pointing to a size-based threshold rather than a published, hard rule (Cult of Mac, last week; iThinkDiff, last week). Apple hasn't published a specific size or pixel threshold separating a marked edit from an unmarked one, so treat this as an early testing pattern, not documented behavior.

Ordinary camera photos sitting untouched in a library aren't automatically watermarked just by existing there. Apple notes that genuine photos already in a library only pick up the mark once they're run through Clean Up, Extend, or Spatial Reframing (9to5Mac, this week).

How to check a SynthID watermark

There's no scanner built into Apple Photos. Apple itself points users to an external tool run by Google (9to5Mac, this week).

The process:

• Go to synthid.com
• Select "Upload content"
• Choose the image to check
• Let Google's SynthID Detector scan it and return a result

Google opened this detector to the public last month, so no developer account or special access is needed to run a check (iThinkDiff, last week).

Do not upload sensitive images unless comfortable sending them to Google's servers for the scan. The moment a photo goes to synthid.com, it's left Apple's ecosystem entirely.

What a detector result can and cannot tell you

A SynthID Detector result isn't a verdict on whether a photo is real. It only answers one narrow question: did the scan find Google's watermark pattern embedded in the pixels?

• Detected: A SynthID watermark was found. That confirms the image passed through a SynthID-compatible generative tool, such as Image Playground, Image Wand, or one of the covered Photos edits (9to5Mac, this week).
• Not detected: This proves nothing about authenticity by itself. The image could be genuinely untouched, lightly edited in a way too small to trigger the mark, regenerated after the fact, or made with an AI tool outside Google's detection network entirely (Cult of Mac, last week).
• Not established either way: A result never confirms who made an image, when it was made, or whether the scene it depicts is genuine. SynthID flags generative-tool involvement; it doesn't verify the underlying content.

That middle category causes most of the confusion. The detector only recognizes watermarks from Google and its partner tools, so a photo built with an unrelated AI generator can come back clean even though it's entirely synthetic (Cult of Mac, last week).

Can you see or remove the watermark, and how durable is it?

The mark is never visible in Photos, and it's never visible to anyone viewing a shared image. It only becomes evident through the detector (9to5Mac, this week).

Because the watermark lives in pixels rather than metadata, Google says it's designed to resist common transformations like cropping, filters, and compression without affecting how the image looks (Cult of Mac, last week). Some reporting describes it as virtually impossible to remove short of cropping a photo down drastically or regenerating it entirely in a tool that doesn't support SynthID (iDropNews, last week). That durability claim comes from secondary reporting rather than independent adversarial testing, so treat it as a directional description, not a guarantee.

What to do now

iOS 27.2 brings SynthID watermarking to Photos and Image Playground, with Image Wand folded into the same always-on coverage. Extend and Spatial Reframing reliably add the mark based on current testing, while Clean Up only sometimes does, depending on how much of the image actually changed (Apple Support, last week; iThinkDiff, last week).

Treat the feature as a provenance signal, not a truth detector. It flags whether Google-recognized generative AI touched an image; a clean result is not proof the image is real (Cult of Mac, last week).

Update to iOS 27.2 if the watermarking feature matters, since the release also brings broader Siri improvements and security fixes alongside the Photos changes (iDropNews, last week). Don't treat a "no watermark" result as proof an image is untouched. And if provenance genuinely matters before sharing or trusting a photo, run it through Google's SynthID Detector at synthid.com instead of judging it by how it looks.