Google opened its SynthID Detector to everyone on October 7, 2026. The free website, synthid.com, lets anyone upload an image, video or audio file and check whether it carries Google’s hidden AI watermark. Until then, the tool was limited to selected journalists, media professionals and researchers, TechCrunch reported.
SynthID is an invisible marker that Google introduced in 2023 and adds to media made with its AI tools, including Nano Banana, Veo, Lyria and Gemini. OpenAI, Nvidia and Kakao also support it, and Google has said Apple support is coming. The detector looks for that marker. It does not judge whether a picture looks fake.
How to check a file in four steps
- Open synthid.com.
- Sign in with a Google, Apple or OpenAI account. The site will not scan anything without a login, Engadget reported.
- Upload your file. It accepts common image types (JPG, PNG, WEBP, HEIC and others), videos in MP4, MOV or WEBM, and audio in MP3, WAV, M4A and similar formats.
- Wait a few seconds for the result.
The public site gives a simple yes or no, according to Ars Technica’s report. A positive result is worded along the lines of “made or edited with Google AI,” The Next Web reported. Google told Ars Technica that each user gets a daily quota of roughly 10 checks. Google engineers said the limit is meant to stop people from using the tool to build ways of removing the watermark.
A clean result does not make a file real
The detector only finds watermarks from companies that use SynthID. Google’s own site says it is not a general AI detector, so an image from a tool that uses another watermark, or none, will come back clean. DPReview tested an image edited with Apple’s AI tools and the detector found nothing, because Apple does not yet add SynthID to its edits.
A positive result also cannot say how much of a file is AI. A fully generated image and a real photo with one object erased can both be flagged as “made or edited with Google AI.”
Google says the watermark survives common changes such as cropping, filters and audio compression. Engadget cautioned that watermarks can be stripped in some situations, and Ars Technica noted that someone may eventually find a reliable way to remove SynthID.
So treat a hit as a strong sign that Google’s AI, or a partner’s, was involved. Treat a miss as “unknown,” not “real.”
Gemini and OpenAI offer other ways to check
You can also upload a file to the Gemini app and ask whether it was made by Google AI. Google’s help page says you can check one file at a time, up to 100 MB. For video, Gemini says which parts of the clip carry the watermark. Google has also built the check into Chrome and says people already make about 1 million verification requests a day, TechCrunch reported.
OpenAI runs its own checker for its content. For a quick check on a viral image, video or voice note before you share it, a watermark scan is a fast first step, as long as you remember what a miss means.
