Can you check SynthID yourself?

Yes, for images, video and audio — in Google's own tools, and nowhere else. In October 2026 Google opened its SynthID Detector portal to anyone with a Google, Apple or ChatGPT account: sign in at synthid.com, upload a file, and it reports whether a SynthID watermark is in it. You can also hand an image or video to the Gemini app and ask whether it was made with Google AI.

Text is the hole. Gemini's writing carries a SynthID watermark too, but the public portal checks media only, so if your question is whether a paragraph came out of Gemini, there is still nothing you can run.

No third-party tool can read it

Detecting a watermark generally needs the parameters used to embed it, and Google has not published a detection API anyone can sign up for — only a content detection API previewed with a few named partners. So no independent site can check SynthID for you, including this one. We read what actually travels with the file: credentials and metadata.

A tool advertising SynthID detection is either running files through Google's portal itself or guessing from the pixels.

A clean result proves nothing

The portal only finds marks from companies that have adopted SynthID: Google's models, plus partners including OpenAI, Nvidia and Kakao. Everything else is invisible to it — Midjourney, Stable Diffusion, any open-weight model running on someone's laptop, and every photograph ever taken with a camera.

So "no SynthID detected" means "not made by one of a short list of models". It is not evidence a person made the file. A hit is the more useful direction: it tells you the content came from a model that marks its output. It does not tell you who made the file, or whether it was edited afterwards.

How SynthID works, and what it covers