Check if that file was made by AI

On October 7, Google opened its SynthID Detector to everyone. It’s a free site — synthid.com — where you upload an image, video, or audio file and it tells you whether the file carries an invisible AI watermark. The launch was announced on Google’s Keyword blog by Pushmeet Kohli, VP of Science and Strategic Initiatives at Google DeepMind. The site is global and in English for now.

Until this week, the detector was a limited tool: an early version launched in May 2025 for journalists, media professionals, and researchers. Now anyone can check files that arrive by email, messaging apps, or download — the exact places where no Google product sits between you and the content. Google says it has watermarked more than 180 billion images and videos and 240,000 years of audio with SynthID since 2023, and that verification checks across Search, the Gemini app, and Chrome now top 1 million a day.

It doesn’t just cover Google’s own output. The detector also reads SynthID watermarks from partner companies — OpenAI, NVIDIA, and Kakao — with Apple support described as coming soon. (OpenAI also runs its own verification site; Microsoft and Meta use their own separate watermarking standards.)

Go to synthid.com and agree to the terms. You’ll sign in with a Google, OpenAI, or Apple account — Google caps free use at roughly ten checks per user per day, a limit meant to stop anyone from hammering the detector to train watermark-removal tools. Then drag and drop a file, or select one: images in JPG, PNG, WEBP, HEIC, GIF, and other common formats; video in MP4, MOV, or WEBM; audio in WAV, MP3, OGG, FLAC, AAC, or M4A.

The result is binary: watermark detected, or watermark not detected. For images it checks the whole file; for video and audio it identifies which segments carry the watermark. Google says uploads are processed in real time and the original file is deleted immediately after the check — only a digital signature of the file is kept for 24 hours to enforce the daily quota.

A ‘watermark detected’ result is strong evidence the file came from one of the participating AI models — Google’s Gemini, Nano Banana, Veo, Lyria, Flow, ProducerAI, and Vids, plus watermarked output from OpenAI, NVIDIA, and Kakao. The watermark lives in the content itself — in the pixels of an image, the frequency components of audio — not in metadata that gets stripped, and Google designed it to survive cropping, filters, frame-rate changes, and compression.

‘Watermark not detected’ is not a clean bill of health. The site’s own FAQ puts it plainly: ‘This is not a general AI detector.’ It can only see SynthID watermarks, so anything from a model that doesn’t use SynthID — most open-weight models, plus Microsoft’s and Meta’s separate standards — comes back clean. Heavy editing can degrade the signal, and Google notes rare false positives. Upload the highest-quality original you have, and treat the answer as one data point, not a verdict.

The practical uses are obvious: the viral image in the family group chat, the voice clip that sounds exactly like someone you know, the video you’re about to share. Check it before you forward it. And you don’t always need the website — SynthID verification is now built into the Gemini app, Google Search (including Lens and Circle to Search), and Chrome, which together handle more than a million checks a day.

One more honest limit: this only works if the AI industry keeps watermarking. Google says it is ‘continuously working on extending partnerships,’ but there is still no single checker that catches everything — Microsoft and Meta run their own standards, and their tools have missed content from their own models. Until that changes, the SynthID Detector is the best free check for a large and growing slice of AI-generated media. Not a lie detector for the internet — a watermark reader, and a good one.