Google Opens SynthID Detector to Everyone: A Public Website for Spotting AI Media
Google launched the public SynthID Detector website on October 7 at synthid.com, letting anyone check images, video, and audio for SynthID watermarks across a long list of file formats, with 1 million verification requests already daily.
Google opened SynthID Detector to the public on October 7: a website at synthid.com where anyone can upload images, video, or audio and check whether it carries SynthID, Google’s invisible watermark for AI-generated content. The portal had been available to journalists, media professionals, and researchers since its introduction at Google I/O the previous year; today’s launch makes it a general public utility, and per TechCrunch’s report, the verification API behind it already handles about 1 million requests daily, built into the Gemini app and Google Chrome.
What It Detects, and the Format List Tells the Story
SynthID Detector identifies watermarked images, videos, and audio clips across an unusually complete format list: JPG, PNG, WEBP, AVIF, HEIC, TIFF, and GIF for images; MP4, MOV, and WEBM for video; WAV, MP3, OGG, FLAC, AAC, and M4A for audio. SynthID watermarks are embedded by Google’s own generation models (Nano Banana, Veo, Lyria) and by generation-capable tools including Gemini, Flow, ProducerAI, and Vids, with OpenAI and Nvidia also supporting SynthID and Apple reportedly planning support. The tool answers one narrow question, was this made by a SynthID-embedding system, and the honest scope statement matters: it cannot detect AI content from systems that use different watermarking standards, such as Microsoft’s and Meta’s, and it says nothing about unwatermarked output at all.
The Detection-Accuracy Problem Is Now Public Infrastructure’s Problem
TechCrunch’s report carries the caveat that defines the category: detection tools “are not infallible, often failing to identify content created by their makers’ models,” citing a Reuters report that Meta’s AI image detector often failed even on its own models’ output after cropping or edits. Making detection a public website converts that limitation from a footnote in research papers into a shared civic experience: millions of people will now learn empirically what watermark robustness means, including that edits can degrade detectability and that absence of a SynthID hit proves nothing about non-Google AI content. That second lesson cuts against intuition in a dangerous way, because the public may read “not detected” as “not AI,” when it often means “not Google AI” or “edited after generation.” A privacy study of conversational AI published last week made the same point about transparency tools generally: their trust effects depend on users understanding what a negative result means.
The Provenance Stack Is Assembling in Public
The launch is one piece of a provenance layer assembling across the industry this year: SynthID watermarks shipping in generation tools from Google, OpenAI, and Nvidia, OpenAI running its own verification portal, Apple reportedly planning SynthID support, and identity vendors building trust infrastructure for AI agents. What distinguishes today’s step is the audience: verification moved from professional tooling to a consumer URL, which makes watermark coverage a network-effect business. Every generation tool that adopts SynthID makes the detector more useful; every high-profile fake the detector misses makes the limitation more famous. Google is betting that being the default provenance checker is worth the misses being public.
What to Watch
Three things. First, adoption by non-Google generators: Apple’s rumored support would be the biggest single win, since iPhone-originated media is the largest real-world provenance market. Second, the adversarial response: watermark-stripping tools against SynthID specifically, now that a free public detector makes testing trivial. Third, false-negative incidents: the first widely shared deepfake that synthid.com fails to flag will shape public understanding of provenance tools more than any launch announcement, and given the Meta detector’s documented failures on edited content, it is a matter of when, not whether.
Get Insanely Good at AI
The book for developers who want to understand how AI actually works. LLMs, prompt engineering, RAG, AI agents, and production systems.
Keep Reading
gr.Workflow Turns AI Pipelines Into Deployable APIs
Learn how to model, debug, expose, and deploy multi-step AI pipelines with Gradio Workflow and daggr.
Gemini 4 Neon: Google's $3/$15 middle model leads Opus 5.5 and GPT-6.2 Sol on most of its own benchmarks
Google released Gemini 4 Neon on November 18, a 1M-context model priced at $3 input and $15 output per million tokens that slots between Hydrogen and Carbon. The pricing, the benchmark table, what it changes for developers, and the safety caveats.
Gemini 4 Hydrogen replaces Flash at $1.50/$7.50: faster agents, triple the Terminal-Bench score, and a higher bill per token
Google released Gemini 4 Hydrogen on November 12 as its everyday model for coding and agents. It scores 51 on the Artificial Analysis index, beats Argon, costs twice Flash's current price, and its model card admits more unjustified refusals.
Google ships a Decisions API at $0.10 per million tokens, as Amazon shows a decider cutting a Sonnet bill by 98%
Google's Decisions API, launched November 11 on Gemini Decision 1, answers up to 1,000 yes/no, choice and score questions per request with no output charges. Two days earlier, Amazon engineers published the numbers from replacing Claude Sonnet 4.6 with a self-hosted decision model.
Gemini 4 Carbon reaches the API: $2/$10 until November 30, then double, and a safety report that calls it no material change
Google opened Gemini 4 Carbon to paying developers and AI Ultra subscribers on November 10, with a 1M-token window and a price rise to $4/$20 on November 30. Its safety report carries Argon's risk findings forward, while showing a 12.8-point jump in exploit success.