Google expanded SynthID watermark detection to a public website and rolled out improved verification for AI-generated media
Deployment ● Confirmed 90% confidence first seen
Google launched and expanded SynthID verification through the SynthID.com website, allowing anyone to check images, videos, and audio clips for SynthID invisible watermarks. The rollout also includes improvements to detection across partner systems and integration in products like Gemini and Chrome, while coverage notes the system may miss some content types such as text and may not detect non-SynthID watermark standards.
Google
SynthID
SynthID AI content detector
SynthID Detector
synthid.com
Anthropic
Claude text watermark
Gemini
Gemini
Google Chrome
OpenAI
Decision brief
- What changed
- Google made SynthID watermark verification publicly available at SynthID.com, where users can check images, video, and audio for invisible SynthID marks from Google and some licensed partners. Google also said it improved detection across partner systems and integrated verification into products including Gemini and Chrome.
- Why it matters
- This gives organizations a broadly accessible way to verify whether some media was produced with Google’s watermarking ecosystem, which can support trust, moderation, provenance, and policy workflows. But decision-makers should treat it as a partial control rather than a general AI-content detector, because the coverage says it does not cover text and does not verify non-SynthID watermark standards, so a negative result does not establish authenticity.
- Evidence
- All three cited outlets report that Google opened SynthID detection to the public via SynthID.com for images, video, and audio. TechCrunch and Ars Technica independently confirm the public rollout and broader product integration, while Trending Topics EU adds the reported limitations around text and non-SynthID standards; these points are consistent across the coverage.
- What remains uncertain
- The coverage does not establish detection accuracy rates, false positive/false negative performance, or how well the tool holds up under editing, compression, or adversarial manipulation. It also leaves open how widely partners have adopted SynthID, so enterprises should assume coverage is limited to media created within that watermarking ecosystem unless Google provides clearer scope.
- Monitor next
- Watch for Google to publish independent benchmark data or product documentation on SynthID detector accuracy, supported partner coverage, and robustness under common media transformations.
Analytical support, not advice — assumptions and open questions stated above.