How to Use SynthID Detector: Methods and Limitations for Verifying AI Images and Videos

When you see photos or videos on social media that are unbelievably vivid, you probably wonder at least once: Is this real, or was it made with AI?

Google has released the SynthID Detector, which checks for AI watermarks in images, videos, and audio, to the public. As of the official update on October 29, it is currently available in English.

However, there is something easy to overlook when reading the results. Just because no watermark is detected, it doesn't confirm that the photo was actually taken by a camera.

What does SynthID check?

SynthID is a technology that embeds digital watermarks—imperceptible to the human eye or ear—into certain AI-generated content. The Detector is a tool for finding these markers in images, videos, and audio.

Google's standalone SynthID Detector is designed to verify AI content from Google and select partners. On the other hand, the SynthID verification within the Gemini app, which I will explain below, is currently limited to content created or edited with Google's AI tools. It's important to recognize that the verification scope for these two services is different.

How to verify photos or videos in Gemini

You can also request verification by logging into the Gemini app and attaching a file. Here is how to do it on a PC.

1. Log in to gemini.google.com.

2. Click 'Add file' and upload the image, video, or audio file you want to verify.

3. Ask a question like the one below and review the results.

Was this image generated or edited with Google AI?

Gemini's SynthID check targets content created by Google AI tools. You can also view metadata like 'Content Credentials' for source and edit history in supported files. Note that even if this information is present, it does not independently verify the authenticity of the media.

Limitations to know before uploading

According to official Gemini guidelines, you can check one file at a time, up to 100MB in size. Videos must be under 90 seconds, and audio must be under 1 hour.

There are also daily usage limits. Guidelines suggest roughly 10 images, 10 videos (total 5 mins), and 10 audio files (total 3 hours) per 24-hour period. These figures apply to the Gemini app and should not be assumed to apply to the standalone Detector.

If a file contains someone else's photo or voice, please check your usage rights and privacy considerations before uploading.

If it isn't detected, is it a real photo?

This is the most important part. Even if the result shows 'No watermark,' it could have been made with other AI tools. There is also a possibility that the watermark could become undetectable after multiple edits or compression cycles.

Conversely, even if a watermark is found, do not assume the entire screen was created by AI. According to Google, it may mean that all or part of the image or video has been generated or edited.

It is best to view the results as just one clue and look into the original publisher, the context of the footage, and the initial posting time. Using reverse image search to see if the material has been posted elsewhere is also a good method.

Rather than a magic button for identifying AI-generated content, it is more of a tool you can use for double-checking when reviewing sources. For important information, always try to find the original source before sharing, rather than relying on a single result.