Google opens SynthID Detector to the public to verify AI‑generated content

In a press release issued on 7 October 2026, Google announced that the SynthID Detector, which had previously been limited to professional users since mid‑2025, is now openly accessible to anyone with an internet connection. The service, hosted at the domain synthid.com, lets users upload an image, video or audio file and receive a simple binary indication of whether a SynthID watermark was found or not. This opening marks the first phase of a broader plan to democratise AI‑generated‑content verification tools, a plan that until now had been confined to media teams and corporate partners.

How the SynthID Detector works

To use the detector, a user visits the website synthid.com, selects the file to be examined and signs in with a Google, Apple or OpenAI account, as described in the tool’s user‑flow documentation. The upload process accepts audio formats (WAV, MP3, OGG, FLAC, AAC, M4A), image formats (JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF, GIF) and video formats (MP4, MOV, WEBM). After the submission, the underlying algorithm scans the file for an invisible digital identifier that is embedded at the moment of content creation and returns to the user a straightforward “yes” or “no” answer.

What the SynthID watermark is

SynthID is an invisible digital watermark introduced by Google in 2023, designed to be automatically inserted into media generated by AI models at the instant of creation. According to the company, more than 180 billion images and videos have already been marked with SynthID, and the total duration of audio that carries the watermark is said to amount to roughly 240 000 years – figures that have not yet been independently verified. The watermark does not alter the visual or auditory appearance of the file; instead, it encodes an identifier that can be read by compatible tools such as the SynthID Detector itself.

Coverage and supported formats

Beyond content generated by Google’s own Gemini models, the detector recognises watermarks inserted by partners that have adopted the SynthID standard, notably OpenAI, NVIDIA and Kakao. Apple has been mentioned as a future participant, although no specific launch date has been set. The extensive list of supported formats means that users can verify virtually any common type of digital media, from JPEG photographs to MOV videos and FLAC audio files, thereby enhancing the tool’s usefulness for both everyday consumers and media professionals.

Integration into Google services

Google has embedded SynthID detection across several of its core platforms, including Gemini chat conversations, Search results, the Chrome browser and the Gemini app. The company claims that, together, these services handle more than one million verification queries each day. This integration allows users to receive automatic alerts when they encounter suspicious content, without needing to visit the dedicated website, although the public tool remains the reference point for more detailed manual checks.

Limitations and uncertainties

A negative result from the detector does not constitute proof of authenticity; it merely indicates that no recognizable SynthID watermark was found in the analysed file. The tool’s effectiveness is limited to content that actually contains the watermark; media produced by systems that do not use SynthID will remain undetectable. Google acknowledges that the watermark is not absolutely tamper‑proof and that some services may strip or modify the identifier, thereby reducing detection reliability. To date, the company has not released quantitative metrics such as false‑positive or false‑negative rates, nor are there independent audits confirming the robustness of the technology.

Implications for the public

By opening the SynthID Detector to the general public, Google dramatically expands the potential user base, moving from media professionals to individual consumers. This expansion could push the volume of verification requests beyond the current level of one million per day, placing additional pressure on backend resources. Industry observers note that, while the watermark represents an important step toward a standardised provenance marker, its presence alone does not solve all challenges associated with AI‑generated content verification; complementary methods such as metadata analysis and human audits remain necessary.

With the public release, anyone can now quickly test whether a file contains a SynthID watermark, thereby strengthening transparency and trust in digital content. What changes now is that individual users, who previously lacked access to such specialised tools, can employ a service that was once reserved for experts, potentially raising broader awareness about the origins of AI‑generated media and encouraging other providers to adopt similar marking standards.