Google Opens SynthID AI Detection to the Public
Google's SynthID, previously available primarily as a developer-facing API and embedded in its own products, now has a standalone public website where users can upload an image and receive a read on whether it carries AI generation or manipulation markers. The interface is straightforward - upload, check, read the result - and requires no account or technical setup.
The core mechanism relies on watermarks that Google's image generation tools embed at the point of creation. SynthID can read those watermarks even after common post-processing steps like compression, cropping, and colour adjustment, which has been one of the system's technical selling points since its introduction. The limitation, as Engadget notes, is that SynthID can only reliably detect images that passed through Google's own generation pipeline. Images produced by other tools - Midjourney, Stable Diffusion, DALL-E, Kling, and others - will not carry SynthID watermarks and may not be flagged.
That caveat matters in practice, since the public-facing tool could give a false sense of reassurance if a user interprets a clean result as confirmation that an image is authentic rather than simply as confirmation that it was not made with Google's tools. The site reportedly acknowledges this distinction, though how clearly it is communicated to non-technical users is worth watching.
The launch comes shortly after research published this week found that AI-edited images cause people to form false memories with high confidence, which sharpens the practical argument for accessible detection tools regardless of their current coverage gaps.
