Is That Photo Real? Google's AI Detector Is Now Open to Everyone, But "No Watermark" Doesn't Mean Real

You see a dramatic photo in your feed. A flooded street, a celebrity in a strange place, a "leaked" image that seems too perfect. Is it real, or did an AI make it?

On October 7, Google opened a tool to the public that tries to answer exactly that question. It is called SynthID Detector, and anyone in the world can now use it. Here is what it does, how to use it, and the one mistake almost everyone will make with it.

What SynthID Detector actually checks

Some AI systems quietly stamp an invisible watermark into the images, videos and audio they create. Google calls its watermark SynthID. You cannot see it or hear it, but software can detect it.

SynthID Detector is a website (synthid.com) where you upload a file and it tells you whether that watermark is present. Until now, the tool was mainly available to journalists and researchers. Now it is open to everyone, although the interface is in English only for the moment.

Three things worth knowing

1. It works on images, video and audio

Common image formats like JPEG and PNG are supported, as well as video files such as MP4 and audio files such as MP3. For video and audio, reports say the tool can point to the specific sections where a watermark appears.

2. It is not just for Google's AI

Besides content from Google's own AI models, the tool also covers content from partners including OpenAI, NVIDIA and Kakao. Apple is reportedly expected to join soon.

3. You need to sign in, and there is a daily limit

You sign in with an existing account, upload a file and get a result. There is a cap on how many checks you can run per day, so it is built for individuals rather than mass scanning.

The mistake almost everyone will make

Here is the part that matters most:

"No watermark found" does not mean "real."

• Content made by AI tools that do not use SynthID cannot be detected at all.

• Heavy editing, cropping or compression can weaken the watermark until it is no longer found.

• In rare cases, the tool can flag content that has no watermark.

In other words, a positive result is a strong hint that a supported AI was involved. A negative result tells you very little. Think of it as one clue, not a verdict.

How to actually use it

1. Check before you share. The more shocking an image is, especially during disasters or breaking news, the more worth it is to pause and check.

2. Combine it with old-fashioned checks. Who posted it first? Is any reliable news outlet reporting the same thing? Does a reverse image search show an older original?

3. Talk to your family about it. AI voices and images are getting good enough that "it looked and sounded real" is no longer proof. Older relatives are often targeted by scams that use exactly this.

The bottom line

SynthID Detector is a useful step. For the first time, ordinary people have an easy way to check whether an image, clip or recording carries a major AI watermark. But it only sees what it was built to see. Use it as a first check, then use your judgment.

Have you ever been fooled by an AI image, or caught one before sharing it? I would love to hear your story in the comments.

This story was written with the help of AI tools. Sources: Google announcement (October 7, 2026), Madonomori (Impress Watch) and other reports. Tool details may change, so check synthid.com for the latest information.