SynthID can help identify content that carries a watermark applied by a Google AI system or another system using SynthID. It is not a universal AI detector. A positive result is evidence about the marked content’s provenance, not who made or shared it, whether it is truthful, or whether an entire file was AI-generated. A negative result means only that the checked watermark was not detected; it does not prove human authorship.
What SynthID checks
SynthID is a family of watermarking and detection technologies developed by Google DeepMind. Google says it embeds signals into AI-generated images, audio, text, and video. A compatible detector checks for those signals; it does not test every possible AI system or determine authorship in general. Google describes the marks as imperceptible to people.
For images and video, Google says the watermark is embedded during creation and designed to withstand changes such as cropping, filters, frame-rate changes, and lossy compression. Its audio watermark is designed to survive common changes such as added noise, MP3 compression, and speed changes. These are design goals, not a guarantee that every transformed file will remain detectable.
Text uses a different approach: the watermark is integrated into token generation. A language model assigns probabilities to possible next tokens, and SynthID adjusts those probabilities so a detector can compare the resulting pattern with expected watermarked and unwatermarked patterns. The result is not a visible label embedded in the words.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problems#1 Best Overall
How to interpret a result
| Result | What it supports | What it does not establish |
|---|---|---|
| SynthID detected | The detector found a SynthID signal in all or part of the submitted content. This is evidence that a system applying SynthID generated or edited the marked content. | It does not identify a person, prove that the whole file was generated by AI, or say whether the content is accurate or misleading. |
| SynthID not detected | The detector did not find the particular SynthID signal it checks for. | It does not prove that a person made the content, that it is authentic, or that no AI was involved. The content may come from a system that does not use SynthID, or a transformation may have weakened or removed a detectable signal. |
| Uncertain text result | The detector cannot confidently classify the sample under its configured threshold. | It is not a definitive yes-or-no authorship verdict. |
A positive result can apply only to a segment, frame, or portion of a file. Google says its SynthID Detector can highlight parts more likely to contain a watermark, and Gemini can provide context about content and segments. Do not treat a marked portion as proof that every pixel, word, or second was AI-generated.
SynthID is a provenance signal, not a fact-check or an intent classifier. Google’s technical explanation puts the limit plainly: “SynthID isn’t a silver bullet for identifying AI generated content.”
Rank #2
Why text results need extra caution
Google says SynthID Text works best on longer, varied responses. Short or factual answers leave fewer opportunities to adjust token probabilities without risking accuracy, so detection is less effective. A thorough rewrite or translation into another language can substantially lower detector confidence. Google says some smaller changes, such as mild paraphrasing or changing a few words, may retain robustness, but that is not assured for every sample.
Text detection is probabilistic. Google’s developer documentation describes three states—watermarked, not watermarked, and uncertain—and says thresholds can be adjusted to change false-positive and false-negative behavior. A result therefore depends partly on detector configuration; it should not be presented as certainty independent of that configuration.
Recommended Free Tools
Where verification is available
Google’s product overview says users can upload an image, video, or audio clip to Gemini and ask whether Google AI created or altered it. Google’s May 19, 2026 announcement said verification for those modalities had been added to Gemini, and that Google was expanding verification to Search immediately and Chrome over the following weeks. Those rollout details are time-sensitive; check the current product before relying on a particular interface or availability in your region.
Google also describes a separate SynthID Detector portal and says it is collaborating with journalists and media professionals to test it. That description does not establish general public access, so do not assume the portal is available to everyone.
For developers, Google’s documentation says SynthID Text is open source and includes a production-grade implementation in Hugging Face Transformers v4.46.0 and later, as well as a reference implementation. Those options are for developers; they do not mean every model or third-party detector uses Google’s watermark configuration.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What adoption and usage figures mean
Google’s May 19, 2026 announcement reported that SynthID had been used to watermark over 100 billion images and videos and 60,000 years of audio. Google also reported 50 million global SynthID verification uses in Gemini as of that announcement. These are company-reported deployment and usage figures, not independent measurements of detection accuracy.
Best Value
Google’s May 20, 2025 announcement had reported over 10 billion pieces of content watermarked. That earlier total has a different date and less specific modality description, so it should not be treated as a directly comparable current total.
In May 2026, Google named OpenAI, Kakao, and ElevenLabs as organizations bringing SynthID technology to more of their AI-generated content. This signals expanding adoption, not that every output from those companies—or every model they offer—contains SynthID. Google also announced an AI Content Detection API on the Gemini Enterprise Agent Platform for organizations evaluating content from Google and other popular models. That broader API should not be confused with a SynthID-only check in Gemini.
How SynthID differs from other provenance signals
SynthID is a watermark check. Other approaches answer different questions: Content Credentials can provide provenance metadata, a visible disclosure can label content for viewers, and a classifier estimates whether content is likely AI-generated. These signals are not interchangeable. When comparing tools, check which signal is examined, which media types are covered, whether the tool detects a mark or infers AI origin, whether it can identify a partial segment, how edits may affect it, and whether it is accessible to the public or restricted to testers or an API.
Google’s May 2026 announcement describes OpenAI, Kakao, and ElevenLabs as adopting SynthID technology, while Google’s product overview discusses SynthID alongside Content Credentials and broader detection services: Google DeepMind’s SynthID overview, Google’s SynthID developer documentation, Google’s May 20, 2025 announcement, and Google’s May 19, 2026 announcement.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
