Updated:
Google on October 7 opened its SynthID Detector to the public worldwide in English, expanding access beyond the journalists, media professionals and researchers who had tested the portal. The tool lets users check image, video and audio files for SynthID watermarks associated with Google’s AI systems and participating partners, including OpenAI, Nvidia and Kakao; Apple support is expected, Google said.
The release broadens public access to a verification system built around watermarks embedded in generated media, rather than offering a general-purpose test for whether any file was produced by AI. Google said more than 180 billion images and videos and 240,000 years of audio have been watermarked to date. The company did not give independent validation of those totals.
From professional testing to public access
Google introduced SynthID in 2023, and launched an early version of the detector in 2025 for media professionals. The new global rollout opens the verification portal to all users, while keeping its current interface limited to English, according to the company’s announcement.
Users can submit media files to check for a watermark. TechCrunch reported that the website supports common image formats including JPG, PNG and WEBP, video formats such as MP4 and MOV, and audio formats including WAV and MP3. The portal focuses on images, video and audio, even though Google’s broader SynthID technology also includes watermarking for text.
Partner companies extend the detector’s reach
The expansion reflects a widening group of companies using SynthID. Google named OpenAI, Nvidia and Kakao as partners whose AI-generated media can be checked through the tool, and said Apple would be supported soon. The announcement did not provide a date for Apple’s integration or specify which of its products would be covered.
OpenAI has also described adopting SynthID for images generated through ChatGPT, Codex and its API, alongside its work on content provenance. Google’s portal is distinct from the separate verification features offered by individual companies and platforms. The partner list does not mean the detector can identify all AI-generated media across the internet.
Watermark detection is not a universal AI test
SynthID embeds an imperceptible digital watermark into supported generated content, which detection software can then look for. Google says the watermark is designed to persist through some common changes to images, video and audio. A file without a detected SynthID watermark, however, cannot be taken as proof that it was made by a person: it may have been created using a system that does not use SynthID or altered in a way that affects detection.
Google’s technical documentation describes SynthID detection as probabilistic and notes that detector confidence can fall after certain changes. The public portal announced on October 7 is for image, video and audio verification, so limitations documented for SynthID text detection should not be treated as a complete technical description of how this portal handles every media type.
Existing verification features handle more than one million requests daily
The public site joins existing SynthID verification features in Google Search, the Gemini app and Chrome. Google said those services collectively handle more than one million verification requests each day. It did not disclose how many of those requests concern each product or how frequently checks produce a positive result.
The next stated development is Apple support, which Google described as coming soon without setting a timetable. The company presented the expansion as a way to make media provenance checks more accessible; how widely users will adopt the standalone portal, and how effectively its partner coverage will grow, remain open questions.
