How to Check if an Image Has SynthID

Here's the part most "SynthID checker" pages won't tell you: you can't reliably scan an arbitrary image for SynthID yourself. It's an invisible pixel watermark that only a matching detector can read, and there's no general public detector — Google's own SynthID Detector is access-gated. So the practical question isn't "how do I scan for SynthID," it's "what *can* I check that tells me whether SynthID is likely present?" The answer is provenance metadata and the source of the image.

Let's do this the honest way.

Why you can't just "scan for SynthID"

SynthID is embedded in the pixels as a faint, deliberately imperceptible pattern. Reading it back requires the specific detector that matches how it was applied. Google has kept broad scanning access limited — partly because a public detector would also be a roadmap for defeating the watermark. The upshot for you: there's no trustworthy "drop your image here, get a SynthID yes/no" tool, and any site claiming a definitive verdict is overselling.

What you can actually check

You can't read the watermark, but you *can* read the file's provenance — and that's often the more useful signal anyway:

• C2PA Content Credentials. Many AI tools, including Google's, attach a C2PA manifest naming the generator. A free metadata checker will surface it.
• EXIF and XMP tags. Software fields and editing history sometimes name the AI tool or pipeline.
• JUMBF containers. The binary format C2PA manifests live in inside JPEGs.

If those show a Google AI origin, it's reasonable to assume SynthID is present too, even though you can't read the watermark itself. If the metadata's been stripped, absence of metadata proves nothing — which is exactly why the source matters.

Use the source as your best clue

Since the watermark is tied to *what made the image*, the most reliable indicator is provenance, not pixel analysis:

• Made with a Google AI tool? Assume SynthID is present.
• Made with a tool built on open models that don't embed it? It almost certainly isn't there.
• A real camera photo? No SynthID at all.

This is also why generic "is this AI?" detectors are unreliable as SynthID checks — they guess at AI-ness from visual cues, which is a different question from "does this carry Google's watermark."

Why this matters for AI dating photos

If you're vetting your own AI photos before posting, the takeaway is freeing: stop hunting for a SynthID scanner that doesn't exist, and instead control the input. Know what tool made your photos, strip the readable metadata, and — for the watermark you can't check — use a generator that never adds one.

That's how our AI dating photos sidestep the whole question: open models that don't embed SynthID, plus a clean re-encode. Nothing to scan for, because there's nothing there. For background, see our SynthID guide. And if you're checking *someone else's* photos, remember reverse image search often tells you more than any watermark hunt.

Frequently Asked Questions

Is there a free SynthID checker? Not a reliable public one for the pixel watermark. Google's SynthID Detector is access-gated. You can freely check provenance metadata (C2PA, EXIF), which is the practical proxy.

Can I tell if a photo has SynthID just by looking? No — it's invisible by design. Visual "AI detectors" guess whether an image is AI-made, which is a different and unreliable question from whether it carries SynthID.

What does it mean if a photo has no metadata? Only that the header data was removed or never added. It doesn't tell you anything about a pixel watermark, which is why the source tool is the better clue.

How can I be sure my own photos have no SynthID? Generate them with a tool that runs on open models that don't embed SynthID. Controlling the source is the only dependable way, since you can't scan to confirm.