Signed Content Credentials
Reads the C2PA manifest that Google, OpenAI, Adobe and other tools sign into their images, and validates the signature against the official C2PA trust list.
Free SynthID Detector · Content Credentials · AI metadata
SynthID DetectorCheck if an image was made with Google AI, ChatGPT or other AI tools
Upload an image and this SynthID Detector reads the signed Content Credentials and AI metadata inside the file. You can see who signed it, whether the signature still checks out, whether the image declares AI generation and whether Google or OpenAI recorded an invisible SynthID watermark. Free, no account needed.
Strictly Prohibits Adult/NSFW ContentDrop an image to check it
JPEG, PNG, WebP, AVIF or HEIC · up to 20 MB · use the original download, not a screenshot
No file handy? Try a real sample:
The free check reads the file on our server and discards it. Nothing is stored.
About the SynthID Detector
What is a SynthID Detector?
SynthID is the imperceptible watermark Google DeepMind embeds in images, video, audio and text made with Google AI. Since May 2026 OpenAI applies it to ChatGPT and API images too. A SynthID Detector answers one question: does this file carry that watermark? Google reads the pixel watermark only in its own tools, so this SynthID Detector works from the evidence that travels with the file — the signed Content Credentials that state a SynthID watermark was applied — and links you to Google’s official SynthID Detector for the pixel check.
SynthID Detector checks an image for the provenance that AI image tools attach to their output. It reads C2PA Content Credentials and validates their signatures against the official C2PA trust list, scans EXIF, XMP and PNG metadata for AI generator fingerprints, and tells you when Google or OpenAI declared an invisible SynthID watermark. The free check runs on our server and the image is discarded afterwards. An optional AI visual analysis looks at the pixels themselves when the file carries no provenance.
Declared SynthID watermark
Shows when the signed credentials state that an imperceptible SynthID watermark was applied — the record Google writes into Gemini and Imagen images.
AI metadata fingerprints
Scans EXIF, XMP and PNG text for IPTC AI source types and generator traces from Midjourney, Stable Diffusion, ComfyUI, Firefly, FLUX and more.
AI visual analysis
When a screenshot or re-upload has stripped every tag, an AI model inspects the pixels for generation artifacts and explains what it sees.
How it works
How to use the SynthID Detector
- 01
Upload an image
Drop a JPEG, PNG, WebP, AVIF or HEIC file up to 20 MB. Use the original download, not a screenshot, for the best result.
- 02
Read the provenance report
See who signed the Content Credentials, whether they validate, whether the file declares AI generation and whether a SynthID watermark is declared.
- 03
Go deeper when nothing is found
Run an AI visual analysis, or check the pixel watermark on Google’s official SynthID Detector with the same file.
Four kinds of evidence
What the SynthID Detector checks — and what each result means
Each layer answers a different question. Signed Content Credentials are the strongest evidence when present; metadata is useful but unsigned; the pixel watermark is read only by Google’s own SynthID Detector; and AI visual analysis is an informed estimate for files with no provenance left.
Content Credentials (C2PA)Signed proof
Cryptographically signed by the tool that made the image. Trusted signers include Google and OpenAI. Stripped by screenshots and most social platforms.
AI metadataUnsigned hint
EXIF, XMP and PNG fields written by generators. Strong when present, but anyone can add or remove it.
SynthID pixel watermarkGoogle’s detector
Embedded in the pixels and robust to edits. Only Google’s official SynthID Detector can read it; this site reports whether one is declared.
AI visual analysis1 credit
An AI model’s estimate from visible artifacts. Useful when no provenance is left, never proof on its own.
Coverage
Which AI images the SynthID Detector can identify
We verified the Google, OpenAI and fal.ai rows with original files on October 8, 2026; the other rows follow each company’s published Content Credentials practice. Every record survives only while the file is unchanged — a screenshot or social media upload removes it.
| Source | What the file carries | SynthID declared |
|---|---|---|
| Google Gemini app (Nano Banana), Imagen, Vertex AI | Google-signed C2PA, states “Applied imperceptible SynthID watermark” | Yes — declared in the credentials |
| ChatGPT, Codex and the OpenAI Images API | OpenAI-signed C2PA with a watermarking step | Yes since May 2026 — watermark step recorded |
| Adobe Firefly, Photoshop generative fill | Adobe-signed C2PA, AI source type declared | No — Adobe uses its own credentials |
| Microsoft Designer, Bing Image Creator | Microsoft-signed C2PA | No |
| API platforms such as fal.ai | Platform-signed C2PA, AI source type declared | Depends on the underlying model |
| Midjourney, Stable Diffusion, ComfyUI | Usually none; metadata or PNG text when kept | No |
Independent vs official
This SynthID Detector and Google’s official SynthID Detector
Google’s official SynthID Detector decodes the watermark in the pixels, so it still works after a screenshot or a crop. It does not show the Content Credentials, the signer, the edit history or AI metadata from non-Google tools. This SynthID Detector shows all of that in one report, validates signatures against the C2PA trust list, and points you to the official SynthID Detector when the pixels are the only evidence left.
Credentials gone? Check the pixel watermark with Google, then come back for an AI visual analysis (1 credit).
Open Google’s SynthID DetectorNew to SynthID? Read what SynthID is and how the watermark works, or use the Content Credentials checker to inspect any C2PA-signed photo or design.
Read results carefully
Can a SynthID Detector be wrong?
A positive result from signed Content Credentials is strong evidence: the AI tool itself declared the image as generated and signed that statement. A negative result is weak evidence. Provenance data is removed by screenshots, editing apps, messaging apps and most social networks, and many AI tools never add it. The pixel watermark is more durable but covers only images from companies that use SynthID. The AI visual analysis is an estimate. Use every SynthID Detector result together with the image’s source and context.
FAQ
SynthID Detector FAQ
What is a SynthID Detector?
A SynthID Detector checks whether an image, video or audio file carries SynthID, the imperceptible watermark Google DeepMind embeds in content made with its AI models and, since 2026, by partners such as OpenAI. This SynthID Detector reads the signed Content Credentials and metadata that accompany those files and shows whether a SynthID watermark is declared, with a link to Google’s official detector for the pixel check.
Can this SynthID Detector read the SynthID watermark in the pixels?
No. Google has not released a public API that decodes the pixel watermark, so no independent site can read it reliably. This detector reads what the file declares — signed Content Credentials state when a SynthID watermark was applied — and links you to Google’s official SynthID Detector for the pixel-level check.
Is the SynthID Detector free?
Yes. Uploading an image and reading its Content Credentials, signature status and AI metadata is free and needs no account. Only the optional AI visual analysis uses credits (1 per image).
Which AI images can it identify?
Any image that still carries C2PA Content Credentials or AI metadata. That includes images downloaded directly from Google Gemini (Nano Banana), Imagen, ChatGPT and the OpenAI API, Adobe Firefly, Microsoft Designer and many API platforms, plus Midjourney, Stable Diffusion and ComfyUI files that keep their metadata.
Why does my AI image show “No AI provenance found”?
Content Credentials and metadata live alongside the pixels and are removed by screenshots, re-saving in many editors, messaging apps and most social networks. The SynthID pixel watermark is designed to survive those changes, so check the same file on Google’s official SynthID Detector, or run the AI visual analysis here.
What does “Trusted” mean on a Content Credentials result?
The manifest’s signature is intact, the image still matches what was signed, and the signing certificate chains to the official C2PA trust list. “Valid” means the signature checks out but the signer is not on that list. “Does not match” means the image or manifest was changed after signing.
Does the detector keep my image?
No. The free check reads the file on our server and discards it when the report is returned. For an AI visual analysis, a downscaled copy is sent to our AI processing provider and the report — not the image — is saved to your account.
Can a SynthID Detector prove an image is real?
No tool can prove that. A missing watermark or missing credentials only means none were found. Treat every result as one piece of evidence and combine it with the source, context and reverse image search.
SynthID Detector is an independent service and is not affiliated with Google, Google DeepMind or OpenAI. SynthID is a Google technology. Results describe the provenance data found in a file and are not legal proof of how an image was made.