Google SynthID Detector Opens to Everyone: What It Can Actually Check

What is Google SynthID Detector?

Google SynthID Detector is a public verification tool that checks whether an uploaded media file contains an invisible SynthID watermark associated with supported AI systems. Google made the detector globally available in English on October 7, 2026, expanding access beyond the media professionals and early users who previously had access. Google’s announcement describes the launch and the supported types of media.

The practical purpose of Google SynthID Detector is narrow but useful. A positive result tells you that the detector found a recognized watermark signal in the file. That can provide provenance information when you are deciding whether a media file came from a participating AI system, although it does not answer every question about who created, edited, or shared the material.

Google has used SynthID since 2023 and says it has watermarked more than 180 billion images and videos, along with 240,000 years of audio. Those figures describe Google’s reported watermarking scale, not the number of files that the public detector can identify in every circumstance. Consumers already following Google AI product updates should treat the detector as a verification feature rather than a general-purpose AI assistant.

What files can Google SynthID Detector check?

Google SynthID Detector can check 3 file categories: images, videos, and audio. Google says users can upload a file in any of those categories to look for an invisible SynthID watermark. The service is designed for a direct file check, which means the result depends on the media file you provide rather than on a description of where the file appeared.

Google SynthID Detector does not operate as a universal judgment about online content. The tool searches for a specific type of watermark associated with participating systems, so a result only addresses whether that signal was found. A file can be AI-generated without carrying a supported SynthID watermark, particularly if it came from a system that does not participate in the program.

The distinction matters most for shared clips, screenshots, and audio snippets that lack a clear source. A detected signal is useful evidence of participating AI involvement. An undetected signal is not reliable evidence that the content is original, untouched, or created by a person.

Which AI systems can Google SynthID Detector recognize?

Google SynthID Detector can recognize supported watermark signals from Google and partners including OpenAI, Nvidia, and Kakao. Google also says support for Apple is coming soon. The participating-systems requirement is the central limit readers should understand before relying on the service.

OpenAI said on July 31, 2026, that supported audio created in ChatGPT and the OpenAI API carries SynthID watermarking, while its image provenance work also uses SynthID. OpenAI’s content provenance announcement explains the company’s participation in the watermarking effort. The partnership means some OpenAI-generated files fall within the detector’s supported scope, not that every AI file online is covered.

Google’s planned Apple support should not be confused with the current ability to sign in using an Apple account. The public detector accepts Google, OpenAI, and Apple accounts for sign-in, according to independent testing. Apple watermark detection remains a future support commitment from Google rather than an available detection category as of October 12, 2026.

How does Google SynthID Detector return a result?

Google SynthID Detector returns whether it found a SynthID signal associated with supported AI systems. The result is therefore a watermark check, not a broad analysis that decides whether a file looks artificial. That design reduces the risk of treating a visual impression, a voice style, or an editing artifact as proof.

Google says its verification features across Search, the Gemini app, and Chrome collectively handle more than 1 million requests per day. The company-reported total indicates that verification has become a recurring use case across Google’s products, although Google did not present that figure as a public accuracy rate for SynthID Detector.

For most users, the most sensible interpretation is straightforward. A positive result supports a conclusion that a participating AI system marked the file. A negative result leaves the source unresolved, so users should check the file’s context, original uploader, and any available provenance information before sharing it as verified.

What are the daily limits and sign-in requirements?

Google SynthID Detector requires an account sign-in, and the public tool accepts Google, OpenAI, and Apple accounts. Ars Technica found that Google gives each account an approximate daily quota of 10 image, video, and audio checks. Ars Technica’s testing of the public tool documented the sign-in options and the approximate quota.

The approximate 10-check limit means Google SynthID Detector is best suited to occasional verification rather than high-volume moderation, archival review, or investigative workflows. Users who need to check many files should not assume that repeated uploads will remain available throughout the day, because the published limit is account-based and described as approximate.

Google has not presented the public detector as a replacement for a newsroom verification process, a platform moderation system, or forensic analysis. The practical response is to reserve checks for files where a supported watermark result would meaningfully change how you interpret or share the content.

What can Google SynthID Detector not tell you?

Google SynthID Detector cannot determine whether every unmarked file is human-made. Ars Technica reports that the tool is not a universal detector for all AI-made media, because it looks for SynthID signals from supported systems rather than attempting to classify every possible image, video, or audio file.

Google SynthID Detector also cannot settle authorship, consent, copyright ownership, or the truth of what a file depicts. A watermark result addresses the presence or absence of a supported signal. A detected signal can clarify that participating AI technology was involved, but it does not explain whether the file was edited afterward or whether a claim attached to the media is accurate.

That limitation is especially important for synthetic audio. Readers evaluating AI-generated voice content can use AI voice generator guidance to understand how easily audio can be created, but watermark detection should remain only one part of an authenticity review.

How should you use Google SynthID Detector safely?

Google SynthID Detector should be used only with files you can safely upload and share. The tool requires users to submit an image, video, or audio file for analysis, so sensitive personal media, confidential work material, financial records, and private conversations deserve additional caution even when the purpose is verification.

Use Google SynthID Detector as one evidence source, then preserve the original file and examine where it came from. A detected watermark can support a decision not to treat a file as unedited original material. An absent watermark should prompt more verification, not a declaration that the content is authentic.

Consumers should also separate content verification from account security. AI-generated media can be used in scams, while compromised accounts can expose unrelated personal information. Readers who store sensitive credentials online should use a password manager with clear security controls and enable the strongest available account protections.

Will Google SynthID Detector solve AI misinformation?

Google SynthID Detector will not solve AI misinformation on its own because the detector only identifies supported SynthID watermark signals. The tool can add useful provenance information for files created by participating systems, but it cannot evaluate every AI generator, every edited file, or every false claim presented beside authentic media.

The tool is still a meaningful consumer option because it makes a previously limited verification service available globally in English. Google says its broader verification features receive more than 1 million requests per day, which suggests users are actively looking for ways to assess AI media. The limitation is that watermark adoption must be widespread for the detector to cover a larger share of online content.

The practical approach is to use the result alongside source checks and basic skepticism. A watermark result can answer a specific technical question about supported AI involvement, while context determines whether the content deserves trust.

FAQ

Is Google SynthID Detector free to use?

Google SynthID Detector is publicly available globally in English, and Google announced the consumer launch on October 7, 2026. Google requires sign-in, while Ars Technica found an approximate daily quota of 10 checks per account.

Can Google SynthID Detector detect every AI image?

Google SynthID Detector cannot detect every AI image because it checks for supported SynthID watermark signals rather than classifying all AI-generated media. A result with no detected signal does not prove that an image was made by a person.

Can Google SynthID Detector check AI audio?

Google SynthID Detector can check uploaded audio files for supported invisible SynthID watermarks. OpenAI says supported audio created in ChatGPT and the OpenAI API now carries SynthID watermarking.

Does Google SynthID Detector work with OpenAI content?

Google SynthID Detector supports watermark signals from OpenAI among Google’s listed partners. The result depends on whether the specific file carries a supported SynthID signal.

Should I upload private files to Google SynthID Detector?

Google SynthID Detector should not be used for sensitive private files unless you can safely share the material for analysis. Use the detector for files where a watermark result is useful and where uploading does not create an unnecessary privacy risk.