Quick answer: Google opened SynthID Detector to the public worldwide in English on October 7, 2026. The online service checks uploaded images, video and audio for invisible SynthID watermarks associated with participating AI systems. It is useful for verifying marked media, but an absent watermark does not establish that a file was made by a human.
What changed on October 7?
Google’s detector began as a verification service available to a limited group of journalists and other media professionals. Google’s October 7 announcement makes that service accessible to the public, with an English-language interface at launch. That matters for anyone trying to assess a suspicious photo, clip or audio recording without having access to specialist newsroom tools.
Google says SynthID has already marked more than 180 billion images and videos and the equivalent of 240,000 years of audio. Those are Google’s stated deployment figures, not a measure of how much AI-generated content is circulating online. The company says related verification features in Google Search, Gemini and Chrome collectively receive more than one million checks per day.
How to use Google SynthID Detector
- Open the official SynthID Detector website. Follow any account sign-in or permission instructions the service presents.
- Select the image, video or audio file you want to examine. Choose an original file rather than a screenshot or heavily recompressed copy when possible.
- Allow the service to analyze the upload, then review the reported watermark findings and any information about which portion of the file was identified.
- Compare the result with the source of the media, publication history and independent reporting before drawing a conclusion about authenticity.
Important: Google has confirmed public access and the supported broad media types, but its initial announcement does not establish one universal maximum file size or one processing time for every format. Check the live interface for current technical restrictions.
Which AI companies are included?
Google says the detector can recognize SynthID-marked content from Google and participating partners including OpenAI, NVIDIA and Kakao. Apple is described as joining soon, which is not the same as saying every Apple-generated file can be identified today. Compatibility depends on the use of SynthID watermarking in the content-production workflow, not simply the name of the company that made the model.
| Claim | What the tool can establish |
|---|---|
| SynthID watermark detected | The file contains a detectable marker from a compatible AI-generation or editing process. |
| No SynthID watermark detected | No supported watermark was found in that check; the file could still be AI-generated. |
| A clip looks realistic | Appearance alone is not evidence that the clip is authentic. |
| A social account claims a clip is real | The claim still requires normal source and context verification. |
Can SynthID detect every AI-generated image?
No. SynthID is a watermark detector, not an all-purpose lie detector or universal AI-image classifier. Some synthetic images and recordings were created by tools that did not apply SynthID. Others may have been altered, exported, rescaled or circulated in ways that affect the availability of machine-readable provenance information.
Equally important, detecting a watermark does not tell you whether the surrounding caption is truthful. A real photograph can be used with a false claim; a watermarked graphic can illustrate a real event. A verification result answers a narrower technical question: was a detectable SynthID marker present?
SynthID Detector vs. Gemini image verification
Google already offers watermark-related checks within some of its consumer products. In Gemini, users can ask whether a submitted file was created using Google AI, and Google describes visual and audio-track verification for supported video uploads. The new public portal serves as a dedicated destination for checking image, audio and video files associated with Google and participating partners.
For journalists, researchers and creators, that separation is practical: you can verify media from one place instead of relying only on whether it appears inside a particular app. Users should still treat it as one step in a broader authenticity workflow.
Privacy and best practices when checking files
- Read the detector’s current privacy notices before uploading sensitive or confidential material.
- Use the original downloaded file when possible and retain the source URL, date and context.
- Cross-check suspicious claims with original announcements, eyewitness records or established outlets.
- Do not label someone deceptive solely because a watermark is present; legitimate disclosure and creative use are common.
- If a check is inconclusive, say exactly that rather than announcing that the image is confirmed real.
Frequently asked questions
Is SynthID Detector available worldwide?
Google’s October 7 announcement says it is available globally in English. Availability of every supported partner model and file format may evolve.
Can it identify AI-generated audio?
Yes, audio is among the media types Google says the detector supports, provided compatible SynthID watermarking is present.
Will it detect content from OpenAI?
Google identifies OpenAI as a participating partner, but that should not be interpreted as a promise that every file made by every OpenAI model contains the watermark.
Does a negative result prove a photo is real?
No. It only means this check did not find a supported marker. You still need source verification.
Sources and further reading
Primary source: Google DeepMind’s October 7, 2026 announcement. Background: Google’s original SynthID Detector introduction. For more changes to conversational AI, see AVARIXO’s GPT-6 Intelligent UI overview.
Other Articles
ChatGPT College Planner for Teens: Features and Release Date
