Google's SynthID lets public scan Gemini, OpenAI, NVIDIA, Kakao media for hidden AI watermarks

Google has made its SynthID detector available to the general public, allowing users to verify whether digital media contains imperceptible watermarks indicating AI generation. This release expands access to a tool previously limited to researchers and journalists in 2025, according to Engadget. The service is now accessible via SynthID.com for anyone wishing to check images, video, or audio files for these embedded labels.

The detector scans for specific SynthID watermarks embedded in content created by participating vendors. It currently identifies media generated by Google’s Gemini model, as well as outputs from OpenAI, NVIDIA, and Kakao. Engadget reports that Apple also participates in the SynthID program, but support for detecting its creations in this public tool is still under development. Users simply upload a file to receive a binary result on whether a watermark was found.

Context and Background

Google introduced the SynthID technology in 2023 to add imperceptible labels to AI-generated content. Prior to this public launch, the company offered a limited version of the detector to select professionals in 2025. The new public tool replaces the previous method where users had to ask Gemini itself to verify content, a process that required engaging with the generative AI model directly. This change provides a standalone verification method for those who wish to avoid using generative AI tools for verification purposes.

Who Is Affected

The tool is designed for general users seeking to verify the authenticity of digital media. It addresses the growing difficulty of distinguishing real content from AI-generated images, videos, and songs. However, the detector has significant limitations. Because it only checks for the presence of a specific watermark, it cannot identify media created by AI models that do not use the SynthID system. Engadget notes that this means the tool cannot detect content generated by Meta’s AI tools or other non-participating vendors.

What should you do?

Security and technology professionals should evaluate the SynthID detector as one component of a broader media verification workflow. While the tool offers a straightforward way to check for specific watermarks, it does not provide comprehensive AI detection. Users must understand that a negative result does not prove content is human-made, only that it lacks a SynthID label. Organizations should continue to monitor vendor updates regarding supported models and detection capabilities.

Background: Generative AI

Generative AI examples include text completion, image synthesis and code generation. These systems predict the next likely token based on vast training datasets. Understanding the underlying mechanics helps you identify risks like data leakage and hallucination in your infrastructure.

Read the full guide: Generative AI Examples: Mechanisms, Risks and Control

What to do and how to stay safe: Google

• Verify media sources through multiple channels rather than relying solely on watermark detection tools.
• Check for official SynthID labels in content from known participating vendors like Google and OpenAI.
• Recognize that absence of a watermark does not guarantee human origin or absence of AI manipulation.
• Monitor vendor announcements for updates on supported models and detection accuracy improvements.

Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Which AI models are currently supported by the public SynthID detector?

The detector currently supports content from Google’s Gemini, OpenAI, NVIDIA, and Kakao models, according to Engadget.

Can the SynthID detector identify content created by Meta’s AI tools?

No, the detector cannot identify media from AI models that do not use the SynthID system, such as Meta’s tools.

Is Apple’s content detection available in the public SynthID detector?

Support for detecting Apple’s creations is still in the works, though Apple participates in the SynthID program.