Google Opens SynthID Detector to the Public for Checking AI-Generated Images, Video and Audio

Google has opened its SynthID Detector to everyone worldwide, giving the public a free tool to check whether images, videos and audio files carry invisible watermarks linked to AI-generated content.

The portal went live globally on 7 October 2026, after an earlier period of access limited to journalists, media professionals and researchers. Google DeepMind, the division behind the system, says the tool can identify content made or edited using Google AI and a growing list of partner platforms.

What SynthID Detector Actually Does

SynthID is a watermarking system Google introduced in 2023. It embeds an imperceptible signal into AI-generated media — one that survives basic edits, compression and format changes. The detector portal scans an uploaded file for that signal and reports back whether a watermark was found, and where in the file it appears.

Results can flag a watermark across the whole file or only in specific portions — useful when edited content mixes AI-generated sections with human-made material.

The tool currently checks images in formats including JPG, PNG, BMP, WEBP, AVIF, HEIC and GIF. For video, it supports MP4, MOV and WEBM. Audio files in WAV, MP3, OGG, FLAC, AAC and M4A can also be uploaded.

Who Made the Content the Detector Can Spot?

SynthID Detector isn’t a catch-all AI detector. It looks for one specific thing: a SynthID watermark. That means it can only identify content created with tools that have adopted the standard.

At launch, those partners include OpenAI, NVIDIA and South Korean technology company Kakao. Apple has been described as forthcoming. Google says more than 10 billion images, videos, audio files and text items have been watermarked using SynthID since 2023 — though the public detector currently covers images, video and audio rather than text.

Google announced the detector concept back in May 2025, initially focused on its own AI tools. The October 2026 launch broadens both access and scope.

The Limits of the Tool

Here’s the part that matters most for anyone tempted to treat a clean result as proof of authenticity: a negative result means nothing.

If the detector finds no SynthID watermark, that does not mean the file was made by a human. It simply means the file doesn’t carry a SynthID signal. Content generated by tools that haven’t adopted the standard — and there are many — will return a blank result either way.

The system’s usefulness also depends on watermarks surviving the journey from creation to distribution. Heavy re-encoding, screenshots or certain editing processes could strip or degrade the signal, though Google says SynthID is designed to be resilient to common modifications.

Part of a Wider Push on Synthetic Media

SynthID Detector sits within a broader industry effort to establish provenance standards for AI-generated content — the question of where a piece of media came from and whether it’s been altered.

Misinformation, deepfakes and AI-generated scam material have pushed governments and technology companies to look for practical verification tools. The UK government has been examining synthetic media risks, though no specific UK legislation was linked to this particular launch.

The detector is available in English and free to use. No account is required, according to Google’s published information at launch.

But the tool is a starting point, not a verdict. Experts advise that anyone assessing suspicious media should also consider the source’s history, the context in which it appeared, and other corroborating evidence — not just a single scan result.

What This Means for Kent Residents

Anyone in Kent — whether a journalist at a local paper, a teacher, a council worker or a member of the public — can now upload a suspicious image, video or audio clip to SynthID Detector and get an instant check, free of charge. It could be a useful first step when assessing AI-generated material circulating in scams, misinformation or impersonation attempts online. But the tool has real limits: if a file comes back clean, that’s not a guarantee it’s genuine, especially if it was made with an AI tool that hasn’t adopted SynthID. Kent residents should treat it as one check among several, not a final answer.