SynthID and PhotoCheck, side by side
They answer "was this made by AI?" in different ways. SynthID looks for a mark Google put there on purpose. PhotoCheck looks at the image itself, and at what the file declares.
| SynthID (Google) | PhotoCheck | |
|---|---|---|
| What it reads | An invisible watermark Google embeds when its tools generate content | The image itself (an AI-generation score), plus C2PA, IPTC and embedded generator data |
| Which generators | Google's own: Gemini, Imagen and other Google AI tools | Any: Midjourney, ChatGPT, Stable Diffusion, FLUX, Gemini and the rest |
| A positive result | Very strong evidence the image came from Google AI | A 0–1 probability; declared metadata is close to conclusive |
| A negative result | Not made by Google AI. It may still be AI from another tool | Lower likelihood of AI from any tool. Still not proof |
| Public API for images | No. A partner-only preview, as of November 2026 | Yes. REST API, 5 free calls a day, then from $0.0075 a check |
| Faces and metadata | No | Face detection, EXIF, GPS and C2PA in the same call |
| Where | Gemini app, SynthID Detector portal (Google sign-in) | This page, no sign-in, or one API call from your server |
- Reads SynthID
- No
- Generators covered
- All
- Public API
- Yes
- Free checks
- 20 a day
Which one to use
Use Google's tools when
- You need to know whether Google's AI made a specific image.
- You're checking a few images by hand.
- Images from other generators aren't your concern.
Use PhotoCheck when
- Images can come from any generator.
- You need it in your product, through an API, at volume.
- You also want faces, EXIF and C2PA from the same call.
They also work well together: run PhotoCheck on everything, and send the images it flags as likely AI to Google's tools when it matters whether Google made them.
How to check an image for SynthID
- Open the Gemini app or gemini.google.com, signed in with a Google account.
- Upload the image and ask "Was this created with Google AI?"
- Gemini looks for the SynthID watermark and answers. A "no" only rules out Google's AI.
Google also offers a SynthID Detector portal. Neither is an API you can call from code, so neither can check uploads automatically.
How PhotoCheck does on Gemini images
We tested our detector on images with known provenance, including seven made with Gemini. All seven scored between 0.60 and 0.90, while four stock photographs scored 0.02 to 0.09. That's a small test: it shows the two groups separate cleanly, not how often that holds. More on Gemini images and how accurate detectors are.
The API, in one call
POST /api/v1/ai-detect returns the AI score and the provenance scan; /analyze adds face detection for the same one credit. Try it without a key: 5 calls a day.
curl https://photocheckapi.com/api/v1/analyze \
-H "Authorization: Bearer $PHOTOCHECK_API_KEY" \
-F image=@photo.jpg{
"ai": {
"score": 0.83,
"verdict": "likely_ai",
"label": "Strong signs of AI generation",
"model": "photocheck-ai-1"
},
"metadata": {
"c2pa": { "present": false, "claim_generator": null, "signature_verified": false },
"ai_declared": false,
"signals": []
},
"summary": {
"ai_generated": "likely",
"text": "Strong signs of AI generation (score 0.83)."
}
}API reference, Python quickstart, pricing. Comparing other tools? See all comparisons.
SynthID questions
Can PhotoCheck detect SynthID?
No. SynthID is Google's proprietary watermark and only Google's tools can read it. PhotoCheck answers the same question a different way: it scores the image for signs of AI generation and reads the provenance labels generators attach, for images from any tool.
Is there a SynthID API for images?
Not a public one, as of November 2026. Google's SynthID Detector is a web portal, and its content detection API is in a preview limited to selected partners. If you need image AI detection from code today, PhotoCheck's REST API is available to anyone.
How do I check an image for SynthID?
Upload it to the Gemini app (or gemini.google.com) and ask whether it was created with Google AI, or use Google's SynthID Detector portal. Both only recognise content made with Google's own AI.
If an image has no SynthID, is it real?
No. Missing SynthID only means Google's AI didn't make it, or that the watermark didn't survive. Midjourney, ChatGPT, Stable Diffusion and others don't use SynthID. That gap is what PhotoCheck's score covers.
How well does PhotoCheck do on Gemini images?
In a small internal test, seven Gemini-generated images scored between 0.60 and 0.90, and four stock photographs scored between 0.02 and 0.09. Seven images show the separation exists; they don't give an accuracy rate.
Should I use both?
If you specifically need to know whether Google's AI made an image, use Google's tools: the watermark is stronger evidence than any score. To check images from every generator, at volume, from your own code, use PhotoCheck.
Is SynthID Text the same thing?
No. Google open-sourced SynthID's text watermarking, but that covers text from language models, not images.
Sources
- Google: AI image verification in the Gemini app: How SynthID checks work in Gemini
- Gemini Help: verify AI-generated images, videos and audio: Google's own instructions
- InfoQ: Google previews a SynthID content detection API: The partner-only API preview, May 2026
- C2PA specifications: The Content Credentials standard
- IPTC Digital Source Type: The controlled vocabulary behind trainedAlgorithmicMedia