Google launched access to its SynthID Detector on October 7, 2026, opening the AI content detection tool to all users. According to MacRumors, the service is now available worldwide in English, though according to Engadget, use requires login via a Google, OpenAI, or Apple account.
How the portal checks files
When users upload a file, they are shown whether a SynthID watermark was found, according to t-online and Google. For video and audio, the portal additionally marks the affected sections. The detector does not indicate whether detected content was created entirely with AI or only edited with AI assistance, according to t3n. The watermark in images and videos should remain detectable even after sharing or edits such as cropping, filters, changed frame rates, or lossy compression, according to Google.
Google's partners from OpenAI to Kakao
Support for Apple Intelligence is due "soon," according to Google. The detector would become one of the first tools able to identify images from Apple's Image Playground, according to Engadget's assessment.
Billions of marked images and videos
As of September 30, 2026, more than 100 billion images and videos as well as over 60,000 years of audio had been marked with SynthID, according to Pushmeet Kohli. Google has since marked more than 180 billion images and videos as well as audio with a total length of 240,000 years with SynthID, according to t3n and t-online.
Content created with Anthropic's Claude is not recognized by the detector, according to t3n; it does not check text at all, according to t3n, which points to other tools such as Pangram. According to t-online, the check occasionally flags material without watermarks; Google recommends uploading files in the highest possible quality for reliable results, since heavy compression or editing worsens detection. Additionally, SynthID watermarks can in some cases be removed through editing, according to t3n and Engadget; Google itself concedes, according to t-online, that with enough trial-and-error edits, one might be found that defeats detection.
Account remains a requirement
Users who do not want to register cannot use the detector with a Google, Apple, or ChatGPT account as an alternative, according to Heise Online; a follow-up inquiry by the publication to Google about a possible change remained unanswered at the time of publication. Engadget, however, lists Google, OpenAI, or Apple accounts as usable access points.
One watermark among several standards
Google emphasizes that its website is not a general AI detector and can exclusively identify media with SynthID technology, according to MacRumors. Beyond that, Adobe, Microsoft, and Google have developed the joint standard C2PA, according to t3n; OpenAI has recently begun watermarking text, though largely optionally, according to Heise Online; Microsoft and Meta use their own methods, according to t3n; and open-source models often do not mark their content. No detector currently exists that recognizes all these standards simultaneously, according to t3n, with reference to trendingtopics.eu. Heise Online assesses SynthID's relatively wide market adoption as an advantage of the tool, but warns that watermarks, while among the best available means for detecting AI-generated content, are not completely protected against manipulation and subsequent removal.