How to Verify AI Media with SynthID Detector

Google expanded its SynthID Detector so anyone can check media for invisible watermarks. The public portal at synthid.com now works globally in English and supports watermarks from Google plus partners including OpenAI, NVIDIA, and Kakao, with Apple support planned.

This guide shows the exact steps to upload files, read results, and understand what a detection does and does not prove.

What SynthID Detector Checks

SynthID embeds an imperceptible watermark into images, video, and audio generated or edited by participating AI tools. Google reports it has watermarked more than 180 billion images and videos plus 240,000 years of audio.

The detector looks for those specific signals. A positive result means the file (or part of it) carries a SynthID watermark from a supported source. It does not scan for every possible AI generator, and a negative result does not prove the media is human-made.

Supported partners currently include Google, OpenAI, NVIDIA, and Kakao. Apple is expected soon.

Step-by-Step: Run a Check

1. Open a browser and go to synthid.com.
2. Sign in with a Google, OpenAI, or Apple account. Login is required.
3. Choose Image, Video, or Audio.
4. Upload a file (drag-and-drop or select). Prefer the highest-quality original available; heavy compression can reduce detection accuracy.
5. Wait for the scan. Video and audio results highlight segments that contain a watermark.
6. Read the outcome. A detected watermark names the likely source (for example, Google AI).

Files are limited in size; check the site for current limits (reports mention around 100 MB). Daily usage may also be capped to reduce abuse.

Reading the Results

When SynthID is detected, the portal states that the media was made or edited with a participating AI tool and notes that further edits may have occurred afterward. For video and audio it can mark specific time ranges.

No detection means no supported SynthID signal was found. The content could still be AI-generated by a non-participating tool, or the watermark may have been removed or degraded.

Always treat the result as one piece of evidence, not a final verdict.

Practical Tips

• Test the highest-resolution version you have.
• Record the date and the exact result if you need it for documentation.
• Do not upload confidential or unreleased client work without permission.
• Combine the check with other signals: C2PA content credentials, reverse image search, and context about the source.
• Remember built-in checks already exist in Search, Chrome, and the Gemini app; the portal is an extra dedicated tool.

For more background on earlier SynthID workflows, see our previous guide at /blog/how-to-check-ai-content-with-synthid-detector/.

Limitations to Keep in Mind

The detector only finds watermarks from partners that have implemented SynthID. It will not catch every AI image, video, or audio clip on the internet. Editing, compression, cropping, and re-encoding can sometimes weaken or remove the signal. Google has limited daily checks partly to make systematic removal harder.

Privacy note: you must log in, and uploaded files are processed by the service. Review the current terms on the site before regular use.

Watch a Walkthrough

Conclusion

The public SynthID Detector portal gives a practical way to check supported media for AI watermarks from Google and its partners. Use it as one verification step alongside other checks, and always interpret a “not detected” result carefully. As more partners join and the tool evolves, it becomes more useful for journalists, creators, and anyone who needs quick provenance signals.

Sources: Google DeepMind blog post (October 7, 2026), Ars Technica reporting, and the official synthid.com portal.