Overview
- Google made the SynthID detector publicly available on Wednesday, October 7, 2026, extending access from registered media professionals to any user worldwide.
- The detector searches for Google's invisible SynthID watermark and related provenance signals embedded in images, video and audio so machines can flag files created by participating models.
- Google says it has embedded SynthID in more than 180 billion files since 2023 and the detector recognizes marks from Google models and partners including OpenAI, Nvidia, Kakao and soon Apple.
- The tool cannot reliably identify all AI content because many generators do not embed watermarks and metadata can be removed or corrupted, so a negative result is not proof the file was human-made.
- For everyday users and journalists the detector offers a practical provenance check but it requires a Google, Apple or ChatGPT account and should be used alongside other standards and tools such as C2PA and OpenAI’s verifier.