A fake image is already on the timeline by the time most people argue about it. Pushmeet Kohli, vice president of science and strategic initiatives at Google DeepMind, wrote on October 7 that the SynthID Detector is now open to everyone, in English, worldwide. Last year the same portal sat on a waitlist for journalists and media staff. The address is synthid.com.
You upload a file. The site looks for SynthID, the invisible mark Google has been writing into images, video, and audio since 2023. Google says it has marked more than 180 billion images and videos, and 240,000 years of audio. Checks already built into Search, the Gemini app, and Chrome regularly handle more than 1 million requests a day. The new site is the version you can open without starting in one of those products. Coverage in Engadget reported that the page asks you to sign in with a Google, Apple, or OpenAI account.
A Pixel camera file is the comparison I want
I shoot on a Google Pixel. The camera uses on-device processing to clean up light, faces, and noise, so the file is not a raw sensor dump. It is still a photograph I took. When a post claims a picture is real, that is the file I want next to a picture made in Gemini. A photograph that never went through Gemini, Veo, Lyria, or a partner generator should come back without a SynthID mark. A generated file from a partner should light up the parts of the frame or the clip that carry the mark.
Google names the partners on the October 7 note: OpenAI, NVIDIA, Kakao, and soon Apple. A hit is evidence that one of those systems wrote the mark. A clean result means the check did not find a supported mark. Someone can still have made the file in a tool that does not use SynthID, or stripped the mark later. Use the upload when a fake image is already circulating. Do not treat a blank result as a birth certificate.
Voice is the check people should run
Cloned voices are already in fraud calls and in clips passed around as interviews. The public site takes audio, which is the format I would try first on a suspicious recording. Google says the portal can flag audio from Google and the named partners. If a clip was built in a tool outside that list, the upload will not name the studio that made it.
On the Pixel there is a separate warning for live calls. Google's Phone app help page says Scam Detection is off until you turn it on, runs on the phone, and does not send the conversation audio or the transcript to Google. On Pixel 9 and later, Google says the warning uses Gemini Nano on the device. In the United States it is available on Pixel 6 and later. If the call looks like someone trying to get money, the phone warns you during the call. That is a different product from synthid.com, and it covers the moment the recording does not yet exist.
Text edits are a harder call
The October 7 expansion is images, video, and audio. I would not expect a public text upload to settle authorship. People already use these tools to remove typos and fix grammar. A paragraph that started as a person's note and left the editor cleaner is ordinary work now. A watermark on that kind of edit would flag routine cleanup along with a fabricated quote. Image and voice files still have a sender, a frame, and a clip you can hold up next to the claim.
Before a customer note or a fraud case treats a posted image or a voice clip as evidence, ask whether anyone uploaded the file to synthid.com, and which account the page required. Then ask whether the live-call warning on company phones is turned on. The site will not grade a press quote that was only cleaned for grammar.
Works Cited
Bell, Karissa. "Google's AI Detection Website Is Now Available." Engadget, 7 Oct. 2026, www.engadget.com/2279565/google-synth-id-detector-ai-detection-website-is-now-available/.
Google. "Use Scam Detection." Phone App Help, support.google.com/phoneapp/answer/15654065. Accessed 7 Oct. 2026.
Kohli, Pushmeet. "Google Expands SynthID Detector for AI Content." Google Blog, 7 Oct. 2026, blog.google/innovation-and-ai/models-and-research/google-deepmind/synth-id-ai-content/.
