Is That Image, Audio, or Video Made by AI? Google's AI Detector Will Tell You

Google opened its SynthID Detector to all users, letting anyone sign in with a Google, Apple, or OpenAI account to check whether an image, audio, or video was generated by Google's AI systems or SynthID partners including OpenAI, Nvidia, and Kakao. Google says it has watermarked over 180 billion images and videos and over 240,000 hours of audio, and the detector's terms of use state that a digital signature of uploaded media is stored for 24 hours to enforce daily quotas and defend against automated attacks before being deleted automatically. Apple will join the program soon, after which Apple Intelligence outputs may also be flagged, while OpenAI said earlier this week it would begin watermarking ChatGPT and Codex text in the EU over the coming weeks and Anthropic did the same in August.

Google is making its AI detection tool, SynthID Detector, https://synthid.com/ available to all users starting today. SynthID has been around for some time. Google uses it to watermark outputs generated by its AI systems, making it easier for end users to verify whether an image, audio, or video was created by Google's AI tools or a SynthID partner like OpenAI, Nvidia, and Kakao. An early version https://blog.google/innovation-and-ai/products/google-synthid-ai-content-detector/ of the SynthID Detector was made available to journalists, media professionals, and researchers via a waitlist last year. Now, it’s open for anyone https://blog.google/innovation-and-ai/models-and-research/google-deepmind/synth-id-ai-content/ ; just sign in with a Google, Apple, or OpenAI account. Apple will join the program soon, after which all audio, video, and images generated by Apple Intelligence https://au.pcmag.com/ai/120222/i-tested-every-apple-intelligence-feature-heres-whats-actually-worth-using may also get flagged by the detector. For now, it can catch outputs from ChatGPT https://au.pcmag.com/ai/101165/chatgpt and Gemini https://au.pcmag.com/ai/101167/google-gemini . I gave the SynthID Detector a quick try by uploading a kitchen redesign mockup I had recently generated using ChatGPT. It correctly flagged the source, both when I uploaded the output and a screenshot of the same. “The watermark doesn’t change the image or video quality,” Google says https://deepmind.google/models/synthid/ . "It’s added the moment content is created, and designed to stand up to modifications like cropping, adding filters, changing frame rates, or lossy compression.” The detector’s terms of use https://synthid.com/terms suggest that Google will store a digital signature of your media for 24 hours to enforce daily usage quotas and “defend the service against automated attacks.” Google adds that it deletes the signatures automatically after 24 hours. So far, Google has watermarked over 180 billion images and videos, as well as over 240,000 hours of audio content. Microsoft also watermarks https://support.microsoft.com/en-us/privacy/watermarks-on-content-generated-or-altered-by-using-ai-in-microsoft-365-at-work-or-school its AI audio, video, and image outputs, while Meta has its own AI image detector https://www.meta.ai/identification/ . As far as AI text detection https://au.pcmag.com/help/99874/5-ways-to-detect-chatgpt-written-text is concerned, there are several third-party options available, while OpenAI said earlier this week that it would begin watermarking text https://au.pcmag.com/ai/120249/openai-to-add-watermarks-on-chatgpts-text-outputs-but-only-in-the-eu generated by ChatGPT and Codex in the EU over the coming weeks. Anthropic did the same https://au.pcmag.com/ai/119320/anthropics-claude-will-now-add-invisible-watermarks-to-text-image-outputs in August. Disclosure: Ziff Davis, PCMag's parent company, filed a lawsuit against OpenAI in April 2025, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.