What happened
Google opened its SynthID Detector to everyone globally in English on October 7, 2026, announcing the move on The Keyword with a post by Pushmeet Kohli, VP of Science and Strategic Initiatives at Google DeepMind. The portal at synthid.com lets anyone upload an image, video, or audio file and check whether it carries an AI watermark — no journalist credentials, no waitlist.
Until now, the Detector was limited to a test group of journalists, media professionals, and researchers who received access after its announcement at Google I/O 2025. Verification features were also quietly built into Search, the Gemini app, and Chrome earlier in 2026 — Google says those handle over 1 million verification requests daily.
Key details
How it works. SynthID embeds an imperceptible digital watermark into AI-generated content at creation time. The Detector scans uploads for that hidden signal rather than guessing from visual patterns like conventional "AI detectors." For video and audio, it can also indicate where in the file a watermark appears.
Scale. Google reports that since launching SynthID in 2023, it has watermarked more than 180 billion images and videos, plus the equivalent of 240,000 years of audio content.
Coverage. The Detector identifies media generated with AI from Google and partners OpenAI, NVIDIA, and Kakao — with Apple support expected later this year. SynthID watermarks are applied to output from Google's Nano, Banana, Veo, and Lyria models, and generation tools including Gemini, Flow, ProducerAI, and Vids. OpenAI, which adopted SynthID support for audio in July 2026, runs its own verification site as well. Microsoft and Meta maintain their own separate watermarking standards.
Supported formats. Images: JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF, GIF. Video: MP4, MOV, WEBM. Audio: WAV, MP3, OGG, FLAC, AAC, M4A.
Important limitations. Google's own FAQ states: "This is not a general AI detector." It can only identify media from companies that have adopted SynthID technology. It cannot reliably distinguish AI-created from AI-edited media, it requires signing in to use, and a missing watermark does not prove a file is genuine — the content may come from a tool that doesn't use SynthID, or the signal may have degraded through heavy editing. The public site reportedly limits users to roughly 10 checks per day, which Google says is intended to make it harder to use repeated checks to develop watermark-removal techniques.
Why it matters for founders and investors
Content provenance just became consumer infrastructure. For founders in media, trust & safety, advertising, insurance, or legal tech, a free global verification portal changes product design: "check this file for an AI watermark" is now a feature anyone can call rather than a specialist capability to build. The smart move is not to replicate detection but to build workflows on top of it — provenance-aware publishing pipelines, evidence chains for legal proceedings, or brand-safety screening for ad platforms.
For investors, the coverage map is the real story. OpenAI, NVIDIA, and Kakao backing SynthID, with Apple expected, creates a genuine industry-standard coalition — but Microsoft and Meta are still running their own standards. Fragmentation is the risk: a startup whose provenance strategy depends on one watermarking standard is exposed. The winners in this space will be agnostic layers that check across standards and combine watermark results with metadata and source analysis.
There is also a compliance angle worth pricing into 2026-2027 planning. As the EU AI Act's transparency obligations and US state-level disclosure rules tighten, the ability to demonstrate that AI-generated content is labeled and verifiable is moving from nice-to-have to a legal requirement. Free global verification lowers the cost of compliance — but only for content covered by the coalition. Anything outside it remains a gap, and that gap is where trust-and-safety startups can still differentiate.
Angeled.com connects founders with investors.
Read Next
OpenAI Releases Codex Security as Open-Source CLI and SDK
Databricks Unveils Always-On Genie, an AI Agent That Stops Waiting to Be Asked
DeepMind Launches AlphaGenome 2, a DNA Model That Reads Genomes Cell by Cell
ElevenLabs Unveils 13 New Products and Upgrades at Its New York Summit
Google Launches Gemini Decision 1, a Model That Answers Instead of Writing
Runway Open-Sources Praxis-1, Its Robot Model That Learned by Watching Video



