What is SynthID?

SynthID is an invisible watermark that Google, and partner companies, add to media made with their AI tools. You cannot see it, and only the detectors run by Google and its partners can read it. On October 7, 2026, Google opened its SynthID Detector to everyone. This page explains what that detector checks, what a result means, and why New Watermark links to it instead of running the check for you.

What Google says the detector does

In a post dated October 7, 2026, Google said the SynthID Detector is available globally in English from that day. Google says anyone can use it to check whether an image, video or audio file was made with AI from Google or its partners, including OpenAI, NVIDIA and Kakao, and, it says, soon Apple. Before that, Google's May 20, 2025 announcement offered the portal through a waitlist for journalists, media professionals and researchers.

The detector's own page, synthid.com, is plain about its limits. It says the tool is not a general AI detector and can only find media from the companies it lists. It accepts images in JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF and GIF formats, plus video (MP4, MOV, WEBM) and audio (WAV, MP3, OGG, FLAC, AAC, M4A). It does not read text.

Which images carry SynthID

The detector covers media from the listed companies, and not every file from those companies carries the watermark. These are the cases we can source:

Google's Gemini API. Google's documentation says all images generated through the Gemini API carry SynthID.
Imagen on Vertex AI. Google's documentation shows a watermark setting that is on by default, and the watermark is SynthID.
OpenAI. It is reported that OpenAI announced on May 19, 2026 that it adds SynthID to images from ChatGPT, Codex and its API. OpenAI's Verify page checks for it.
NVIDIA, Kakao and Apple. Google names them as partners. We found no public description of what each adds to images today.

The Gemini app is a special case. It is reported that since August 14, 2026 users can turn off its visible sparkle mark in Settings, under Media Watermark, and that turning it off leaves SynthID and Content Credentials in place. See visible AI watermarks.

How to check an image with SynthID Detector

Open synthid.com. It is Google's site, not ours.
Sign in. Android Authority reported that Google, Apple and OpenAI accounts work for sign-in.
Upload the file. Use the file as you received it; we cannot tell you how well the watermark survives edits.
Read the result, which is either "Watermark detected" or "Watermark not detected".

The image is uploaded to Google. According to the terms dialog on synthid.com, uploads are deleted immediately, and a digital signature of the media, mapped to your user ID, is kept for 24 hours to enforce usage quotas. Quotas exist; we could not find the number.

What a result means

Watermark detected. Google's detector found a SynthID watermark in the file. Google also says that in very rare cases SynthID can falsely trigger on content that has no watermark.
Watermark not detected. Google says this does not rule out AI. The file may come from a company the detector does not cover, it may never have carried the watermark, or editing may have removed it.

SynthID Detector answers one narrow question about one family of watermarks. It does not say who made an image, and "not detected" is not evidence that an image is real.

Why we link to it instead of running it

Google's announcement describes no API, SDK or embed for the detector. Its terms bar bypassing the website to submit requests programmatically, or using the site as an automated API. A separate Google Cloud "AI Content Detection API" was reported by InfoQ in May 2026 to be in preview for selected partners, including Shutterstock, Snap, Fox Sports and Canva. A page like ours can link to the detector but cannot run it on your behalf.

So every result on our checker carries the line "Not checked here: Google's SynthID watermark", and the results page links to Google's detector. We never pre-fill anything and never pass your image on. If you follow the link, you upload the file to Google yourself.

What we check instead

Where an image carries them, our checker reads Content Credentials, with real signature and trust checks, and AI declarations in metadata. Those signals are different from SynthID and can exist without it, or the other way round. Google's documentation describes both for its Gemini image models. Run both checks when the answer matters. For files from ChatGPT, OpenAI Verify covers Content Credentials and SynthID in one place. What are Content Credentials? explains the signed record, and How to tell if an image is AI puts all the signals side by side.

Frequently asked questions

What is SynthID?: An invisible watermark that Google and partner companies add to media made with their AI tools. Only their detectors can read it.
Can I check SynthID without signing in?: Not that we know of. Sign-in is required on synthid.com. Reports say it accepts Google, Apple or OpenAI accounts.
Does SynthID Detector work on any AI image?: No. It finds media from Google and the partners it lists. A "not detected" result does not rule out AI.
Does New Watermark check SynthID?: No. We link to Google's detector. Our checker reads signed records, metadata and one open invisible watermark, and it says on every result that SynthID was not checked.
Is there an API for SynthID Detector?: Google's announcement describes none, and its terms bar automated use of the website. A Google Cloud API was reported in preview for selected partners.
Does Google keep the images I upload?: Google's terms dialog says uploads are deleted immediately. It also says a digital signature of the media, linked to your user ID, is kept for 24 hours for quota purposes.