From Invite-Only to Everyone
Google has officially opened its SynthID Detector to the public. Announced on October 7, 2026, the free website lets anyone verify whether an image, video, or audio clip was generated using AI. The tool was first revealed at Google I/O in May 2025 as a limited verification portal for journalists, media professionals, and researchers — now, after a waitlist phase, it is open to everyone.
The timing could not be better. AI-generated deepfakes are flooding social media, from fake celebrity videos to manipulated news clips, and most people have no reliable way to tell what is real. Google says its detector is now being used to make 1 million verification requests every single day — and that number will only grow now that the gates are open.
How SynthID Works
SynthID is Google's invisible watermarking technology, first introduced in 2023. When Google's AI models generate media, they embed an imperceptible digital watermark directly into the pixels, waveforms, or text tokens. Human eyes cannot see or hear it, but the SynthID Detector can read it.
When you upload a file to the detector, it scans the media and highlights the regions most likely to contain a watermark. Models covered include Google's Nano Banana, Veo, and Lyria, plus tools with generation capabilities like Gemini, Flow, ProducerAI, and Vids. SynthID watermarking has also expanded to text, audio, and video — and Google has open-sourced its SynthID text watermarking so others can adopt it.
What the Detector Supports
The public site accepts a wide range of file formats:
- Images: JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF, and GIF
- Videos: MP4, MOV, and WEBM
- Audio: WAV, MP3, OGG, FLAC, AAC, and M4A
Simply upload a file and the detector tells you whether it carries a SynthID watermark indicating AI generation. Verification is also baked directly into the Gemini app and Google Chrome, so checks can happen without visiting the website at all.
A Growing Industry Standard
SynthID is becoming bigger than Google alone. OpenAI, Nvidia, and Kakao also support SynthID watermarking, and OpenAI maintains its own verification site for checking content. Apple is expected to add support soon. Google also partnered with content-verification firm GetReal Security, and in May 2026 launched an AI Content Detection API on Google Cloud's Gemini Enterprise Agent Platform for business users.
Microsoft and Meta have their own separate watermarking and verification standards — but Google's bet is that a shared, open technology will cover more ground.
The Honest Limitation: It Is Not Infallible
Here is the important caveat: these tools are not perfect. Independent testing has shown that AI-detection tools can fail to identify content created by their own makers' models, especially after editing, compression, cropping, or re-recording. A clean result from the SynthID Detector does not guarantee a file is authentic — it only confirms whether a recognized watermark is present. Screenshots, screen recordings, and heavily edited files can lose the watermark entirely.
Practical tips for staying safe:
- Use the detector as one signal, not the final verdict. Combine it with reverse image search and source checking.
- Check the original source. A video shared on WhatsApp has already been compressed many times — the watermark may be gone even if it was AI-generated.
- Look for inconsistencies. Blurry hands, mismatched lip movement, and unnatural lighting remain useful human tells.
- Verify before you share. If a shocking clip arrives in your family group, spend 30 seconds checking it before forwarding.
Why This Matters for Pakistan
AI-generated misinformation is already being used in Pakistani political and religious discourse, and manipulated videos of public figures have gone viral here multiple times. A free, public verification tool that anyone can use is a genuine step forward for media literacy — especially when elections, financial scams, and viral WhatsApp forwards are involved.
Disclaimer: Detection tools have known limitations and cannot definitively prove whether media is real or fake — always verify through multiple sources.



