SynthID Detector
Check whether an image was made with Google AI — read its Content Credentials and metadata in your browser, run a pixel-level AI scan, then confirm the watermark in Google's own detector.
- Credentials & metadata read in your browser
- Free check — no account needed
- 10 languages
What is a SynthID detector?
A SynthID detector checks a file for signs it was generated by Google AI. SynthIDDetector.org reads the file's Content Credentials (C2PA) manifest and AI metadata in your browser, and can run a pixel-level AI scan — then points you to Google's official detector when the signals suggest a SynthID watermark.
| What it checks | Content Credentials (C2PA), AI metadata, pixel-level AI scan |
|---|---|
| Input | JPG, PNG or WebP image up to 10 MB |
| Output | Per-signal report — credential signer, AI markers, scan verdict |
| Free check | Yes — credentials and metadata, no account needed |
| AI scan | Sign-in required; uses credits |
| Privacy | Credential and metadata checks run locally in your browser |
| Languages | 10 |
How SynthID detection works
SynthID is an invisible watermark — only Google's detector can read it. This tool reads the public signals around it.
- Content Credentials (C2PA)Signed by Google LLC · declares trainedAlgorithmicMediaCredential found
Content Credentials (C2PA)
Google signs Gemini, Nano Banana, Imagen and Veo output with a C2PA manifest — signer “Google LLC”, app “Google Media Processing Services” — that declares AI creation and can state a SynthID watermark is embedded. We read it in your browser.
- AI metadataIPTC DigitalSourceType · XMP CreatorToolAI markers found
AI metadata
Generators also leave softer traces — IPTC DigitalSourceType, XMP CreatorTool, EXIF Software and PNG text chunks. Not proof on their own, but strong context next to a credential.
- Pixel-level AI scanInspects pixels, not metadata — survives screenshotsAI signals detected
Pixel-level AI scan
Metadata dies with the first screenshot or social re-upload. A signed-in pixel scan inspects the image itself and reports an AI probability plus a likely generator family.
The SynthID watermark itself is proprietary — only Google's SynthID Detector can confirm it. When our signals point to Google AI, we send you there to verify.
How to read a result
Each signal answers a different question — combine them before you conclude.
Google-signed credential found
A valid C2PA manifest signed by Google is strong evidence the file came from a Google AI product. The manifest can also declare a SynthID watermark — confirm it in Google's detector.
AI signals without a credential
Metadata markers or a high pixel-scan score suggest AI generation, but cannot prove which model made the image or whether a watermark exists.
Nothing found is not “real”
No credential, no markers, low AI score — that means no signal was detected, not that the image is genuine. Screenshots and re-uploads strip credentials entirely.
Mixed signals
A credential from one tool plus metadata from another usually means the file was edited or re-exported after generation. Read each field separately, don't add them up.
How to check an image
Four decisions, one minute.
- 01
Use the original file
Start from the original export — a screenshot, re-upload or chat attachment usually strips the credentials. Get the file as it left the generator.
- 02
Drop it in the detector
Content Credentials and metadata are read instantly in your browser — free, no account, and the file is never uploaded for these checks.
- 03
Run the AI scan when nothing is embedded
If the file carries no credentials or metadata, sign in and run the pixel-level scan — it still works on screenshots and re-uploads.
- 04
Confirm the watermark with Google
When signals point to Google AI, verify the SynthID watermark itself in Google's official detector or the Gemini app.
Coverage by AI generator
What each source writes into its files — and which signal catches it.
| AI source | Content Credentials | SynthID watermark | Pixel AI scan | Best next step |
|---|---|---|---|---|
| Google Gemini / Nano Banana / Imagen | Signed by Google | Yes | Covered | Confirm in Google's detector |
| Google Veo (video) | Signed by Google | Yes | Video scan planned | Check the video in Google's detector |
| OpenAI ChatGPT images | C2PA manifest | Yes — verify with Google | Covered | Read the credential, then confirm with Google |
| Adobe Firefly | C2PA manifest | No | Covered | The credential is the strongest signal |
| Midjourney | None | No | Covered | Use the pixel AI scan |
| Stable Diffusion / Flux (local) | None | No | Covered | Use the pixel AI scan |
SynthIDDetector.org vs Google's SynthID Detector
Different jobs — use both.
| Capability | Google's SynthID Detector | SynthIDDetector.org |
|---|---|---|
| Reads the SynthID watermark | Yes | No — we link you to it |
| Content Credentials (C2PA) detail | No — watermark only | Yes — in your browser |
| AI metadata fields | No — watermark only | Yes — in your browser |
| Pixel scan for non-watermarked generators | No — watermark only | Yes — after sign-in |
| Works after screenshots / re-uploads | Often — the watermark survives | Pixel scan: yes; credentials: stripped |
| Languages | English | 10 |
| Sign-in required | Google account | None for the free check |
Use cases
Where a quick signal check saves you from a bad call.
Verify a viral photo
A dramatic news or social image is spreading — check for a Google-signed credential before you share or publish it.
Check a profile photo
Dating and marketplace profiles: see whether a photo was generated rather than photographed before you trust it.
Moderate AI content
Trust & safety and marketplace teams: spot-check uploads against your AI-labelling rules.
Classroom and assignments
Teachers and reviewers: check whether a submitted image was generated before you grade or cite it.
SynthID Detector FAQ
Straight answers about what this tool can and cannot see.
Does this tool read the SynthID watermark itself?
No. Only Google's own SynthID Detector can read the watermark. This tool reads the public signals around it — Content Credentials, AI metadata and a pixel-level scan — and links you to Google's detector to confirm.
Can I check Gemini or Nano Banana images?
Yes. Google signs Gemini, Nano Banana, Imagen and Veo output with Content Credentials, which this tool reads in your browser. A valid Google signature is strong evidence the file came from Google AI.
If nothing is found, is the image real?
No. “Nothing found” means no signal was detected — not that the image is genuine. Screenshots and social re-uploads strip credentials and metadata, so treat an empty result as inconclusive and run the pixel scan.
Does SynthID survive screenshots and compression?
The SynthID watermark is designed to survive common edits and often survives screenshots — which is why Google's detector may still find it after the credentials are gone. Our pixel scan works the same way: it reads the pixels, not the metadata.
Is my file uploaded or stored?
The free credential and metadata checks run entirely in your browser — the file is never uploaded. Only the optional pixel AI scan sends the file for analysis; files are not published.
Does it detect ChatGPT, Midjourney or Flux images?
The pixel scan looks for AI generation regardless of the generator and names a likely family when confident. ChatGPT and Adobe Firefly also write C2PA credentials; Midjourney and local Stable Diffusion or Flux write none.
Does it check video or audio?
Video and audio checks are planned for a later version. Google's official detector already accepts video and audio with a Google sign-in.
Why do results differ from Google's detector?
Only Google can read the SynthID watermark. This tool reads complementary public signals — treat our result as a pointer and Google's verdict as the final word on the watermark.
Pricing and credits
Content Credential and metadata checks are free — no account. Credits are used for pixel-level AI scans.
Basic
Pro
Max
Need more flexibility?
View full pricingRelated tools & guides
Narrower checks and explainers built on the same honest signals.
Check your image now
Drop it in — credentials and metadata are read instantly in your browser, free.