iPhone SynthID Watermark: How to Check AI-Edited Photos

Since iOS 27.2, released October 26, 2026, images made in Image Playground and photos changed with larger Apple Intelligence edits carry an invisible SynthID watermark. Your iPhone can’t detect it: there is no setting or label for it in Photos. To check a file, upload the original to Google’s SynthID Detector at synthid.com, but don’t treat a “no watermark” result for an iPhone image as proof it’s real, because Google hasn’t confirmed its tools read Apple’s watermark yet.

This guide covers which iPhone edits are watermarked, how to run a check, what the results mean and the metadata label you can look at instead. It was checked on October 27, 2026 against Apple’s June announcement, Google DeepMind’s October 7 post, Gemini’s help page and MacRumors’ testing. We could not run detector tests on iOS 27.2 images ourselves.

Before you start

You’ll need: The original image file (not a screenshot or a messaging-app copy) and a web browser
Time: About 2 minutes per image
Watch out: Uploading sends the image to Google. Don’t check private photos you wouldn’t want processed there

What changed in iOS 27.2

Apple said in June 2026 that Apple Intelligence edits in Photos would get “a hidden SynthID watermark,” and that Image Playground images would carry the same marker. That arrived with iOS 27.2. SynthID is Google’s watermarking system: it hides a signal in the image pixels rather than in metadata, and Google says it’s designed to survive cropping, filters and compression.

Google says SynthID has been applied to more than 180 billion images and videos. OpenAI, NVIDIA and Kakao also use it, which is why one detector can check images from several companies.

Which iPhone edits get a watermark

Feature	Watermarked?	Basis
Image Playground (new images)	Yes	Apple announcement; MacRumors testing
Extend in Photos	Yes	MacRumors testing
Spatial Reframing in Photos	Yes	MacRumors testing
Clean Up in Photos	Only larger edits; small fixes had none	MacRumors testing
Crop, exposure, filters, Portrait edits	Not described as watermarked	These don’t generate new pixels with AI
Genmoji	Not covered	Neither Apple nor MacRumors addressed it

Apple hasn’t published the size threshold for Clean Up, so a lightly retouched photo may come back clean even though AI touched it.

How to check a photo with SynthID Detector

Get the original file. Screenshots and copies sent through messaging apps are recompressed, which can weaken the watermark. On iPhone, check the photo straight from your Photos library.
Open synthid.com in Safari or any browser. Google opened the detector to everyone on October 7, 2026, in English.
Upload the image. The detector accepts images, video and audio. Third-party reports put the limit at about 10 checks a day.
Read the result, using the table in the next section.

The catch: Google’s October 7 post listed Apple as “coming soon” among SynthID partners, and the Gemini help page says Gemini “can currently only recognize content created by Google AI tools.” MacRumors points readers to the detector for iPhone images, but we couldn’t confirm from Google or Apple that Apple’s watermark is readable today. Test with an image you know came from Image Playground first; if it isn’t flagged, that’s a strong sign the detector doesn’t read Apple’s watermark yet.

Checking with Gemini instead

At gemini.google.com, select the add-files button, choose Upload Files and ask “Is this AI-generated?” Gemini checks for SynthID and Content Credentials. Per Google, it recognizes Google AI content only, so it’s useful for images that may come from Gemini or Google’s image models, not for confirming iPhone edits. The limit is about 10 image checks in a rolling 24 hours, one image at a time, up to 100 MB.

Check the metadata label for Clean Up

Apple also writes edit information into a photo’s metadata. Since iOS 18.1, photos edited with Clean Up are tagged “Modified with Clean Up,” according to AppleInsider. Open the photo in Photos and tap the info (i) button to see the details Photos shows.

Metadata is easy to lose. Screenshots, many social apps and some export settings strip it, so a missing label tells you nothing. The SynthID watermark exists because metadata isn’t reliable.

What a detector result means

Result	What it means	What it doesn’t mean
Watermark detected	AI from a supported company created or edited all or part of the file	How much was changed. A small fill and a fully generated image can look the same
No watermark found	No supported watermark was read	That the image is real. Other AI tools, heavy edits and recompression leave no readable mark
Inconclusive	The signal was too weak to call	Anything either way. Try the original file

Google advises treating a result as one clue among several. For images tied to news or elections, check where the image first appeared and whether reputable outlets carry it. The same skepticism applies to stores and marketplaces, as in our guide to spotting AI-generated games on Steam.

Can you turn the watermark off?

Apple hasn’t documented any setting to turn SynthID off for Image Playground or Photos edits. If you don’t want a watermark, the only route is not using the AI editing tools. The watermark is invisible and Google says it doesn’t affect image quality.

If you haven’t installed the update yet, our iOS 27.2 overview covers the other changes, and iOS 26.7.2 vs iOS 27.2 helps you decide whether to stay on iOS 26.

FAQ

Does the iPhone have a SynthID detector?

No. iOS 27.2 adds the watermark but no way to read it. MacRumors confirmed there’s no detection option on the iPhone.

Do ordinary iPhone photos get a watermark?

No. Based on Apple’s description, the watermark is applied to AI-generated images and AI edits, not to photos straight from the Camera app.

Does the watermark survive screenshots?

Google says SynthID is built to survive cropping, filters and compression, but heavy editing, screenshots and reposts can weaken it. Always test the original when you can.

Can other people see the watermark?

No. It’s invisible to the eye. Anyone with your image can upload it to a detector, though.

What to do next

Generate one test image in Image Playground and upload it to synthid.com. If it’s flagged, you know the detector reads Apple’s watermark and you can rely on it for other iPhone images.

iPhone SynthID watermark: how to check if a photo was AI-edited

What changed in iOS 27.2

Which iPhone edits get a watermark