Google opens SynthID Detector to everyone: images, video and audio can now be checked for an AI watermark
On October 7, 2026 Google made its SynthID Detector available to the public worldwide. The tool at synthid.com runs in English for now. Upload a picture, video or sound file and it tells you whether the content came from AI built by Google or by partners such as OpenAI, NVIDIA and Kakao.
From a limited test for media professionals to a public website
The announcement went up on Google's blog on October 7, 2026, written by Pushmeet Kohli, Google DeepMind's VP of Science and Strategic Initiatives. According to his post, SynthID Detector can now be used from anywhere in the world through an English interface. You upload a file and the tool shows whether it was generated with AI models from Google or from SynthID partners. Those partners currently include OpenAI, NVIDIA and Kakao, and Google says Apple will join them soon.
The tool itself is not brand new. Google says an early version arrived last year to help media professionals verify AI content, and TechCrunch recalls that select journalists, media professionals and researchers got to test it at last year's Google I/O. What changes now is that any user can run the same check. Google positions the site alongside the verification features already built into Search, the Gemini app and Chrome, which it says now regularly handle more than 1 million requests a day.
Which files it accepts and where the watermark comes from
TechCrunch lists the accepted formats. For still images: JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF and GIF. For video: MP4, MOV and WEBM. For sound: WAV, MP3, OGG, FLAC, AAC and M4A. In practice, the social post, short ad video or voiceover file an agency hands over will arrive in one of these formats.
What the tool looks for is the SynthID watermark, an imperceptible signal Google has been embedding in its generated media since 2023. TechCrunch reports that Google's Nano Banana, Veo and Lyria models, along with generation-capable tools such as Gemini, Flow, ProducerAI and Vids, all apply it to their output. The scale is large. In Kohli's words, Google has "watermarked over 180 billion images and videos, along with 240,000 years of audio content" since SynthID launched in 2023.
Using AI is now a claim anyone can test
For advertising and creative production, the shift is this: an image or video made with a Google or partner model now carries a trace anyone can query in a few steps. A brand saying it used no AI in a campaign is making a statement that a client, a competitor or a journalist can test on their own. That moves disclosure of AI-assisted creative away from being optional and closer to being part of brand reputation.
The limits deserve equal attention. Google positions the detector as a check for content made with its own AI and that of SynthID partners. TechCrunch notes that OpenAI runs a separate site of its own for checking content, and that Microsoft and Meta each follow their own watermarking and verification standards. The same report says these tools are not infallible and frequently miss content generated by their own makers' models. So when no watermark turns up, that should not be read as proof a human made the content.
The sources say nothing about Türkiye, so this section is commentary
Neither source contains any information about Türkiye. Google says the tool is available globally in English; there is no separate announcement, local partner or usage figure for Türkiye. This section is therefore not a sourced fact but clearly UNALSOFT's commentary. The unknowns, one by one: no date is given for languages other than English, and Apple support is only described as coming soon, without a date. The sources do not say whether output from generators outside the partner list, such as Midjourney, Runway or Kling, is recognised, whether cropped, edited or re-encoded files are still detected, what file size or daily usage limits apply, or how accurate the tool is and how often it returns false positives.
Our reading: for brands in Türkiye producing images and video with AI, the tool works as a useful second pair of eyes for internal checks, but it is not a certificate on its own. Recording which creative was made with which tool, at which stage, is a sturdier basis than a watermark check after the fact. We looked separately at Google's updated guidance that AI-generated content should be checked by a person before publishing on October 2; that article is independent of this tool.
UNALSOFT's take
We read this less as a detection story and more as a discipline of transparency in creative production. Google adds the watermark automatically to output from models such as Veo and Nano Banana, so an ad image or video made with them carries a trace anyone can query for as long as the watermark is present. The practical consequence is plain: if you use AI-assisted creative, say so to your client and your audience yourself, and keep a record in the project file of who produced what with which tool. That is why on the AI ad films side we treat the production log as part of the delivery. Because the sources carry no data about Türkiye, this article proposes no local legal conclusion.
Can you stand behind the creative you make with AI?
A short conversation is enough to set up a creative process that records which content was made with which tool and discloses it openly.