Google has begun rolling out SynthID watermark detection in its Gemini app, letting users ask whether an image was created by Google’s own AI systems.
From today, tapping the + icon, selecting an image and typing “Was this image generated by AI?” will trigger Gemini to scan for SynthID, an invisible watermark embedded in more than 20 billion pieces of Google-generated content since 2023. The app replies with reasoning and context about the file’s origin regarding whether it was generated by Google’s AI models.
Free and Pro users will continue to see a visible “Gemini glow” watermark on new images, while Ultra subscribers can export clean versions for professional use. Google plans to extend SynthID to audio and video later this year.
During brief tests, Gemini correctly identified an image made with Google’s Nano Banana Pro, but when shown a picture created by ChatGPT — which lacks SynthID — the app could only offer a cautious guess based on visual cues, concluding it was “probably AI-generated” but “not possible to determine” the tool used.
Google acknowledges that universal detection will remain limited until other platforms adopt compatible watermarking.