Promptpulse
artificial inteligenceTelegram

Google Opens a Public Site for Checking Whether Media Is AI-Made

AI Tools·October 8, 2026

Google on Tuesday launched a website that lets anyone check whether an image, video, or audio clip was generated using artificial intelligence. Visitors upload a file and the site looks for SynthID, the invisible watermark Google has been building into content created with its own generative models.

SynthID works by embedding a signal into the content itself, in the pixels of an image, the frames of a video, or the waveform of a sound. Google designed the mark to survive ordinary changes such as resizing, compression, or light editing, so it can still be detected after a file has been shared around. The check is meant to answer one narrow question: was this made with a Google model that applies the watermark?

That narrowness is the main limitation. A file without a SynthID mark is not proof that it is authentic. It could have been made with a different company's tool, or it could be a real photo or recording that was never watermarked at all. A positive result, on the other hand, is a strong signal that the content came from a SynthID-enabled model, and that makes the tool useful for quickly checking suspicious images and clips.

The launch comes as synthetic media has become cheap and convincing. Fabricated photos, cloned voices, and AI-generated video have shown up in political ads, scams, and viral hoaxes, and newsrooms and fact-checkers have been looking for faster ways to verify what they receive. Putting a detection tool on a public website moves that capability out of internal systems and into the hands of reporters, educators, and ordinary users who want to take a quick look before they believe or share something.

Watermarking is one piece of a broader push to label AI-made content, alongside industry standards for provenance metadata and disclosure rules in some jurisdictions. None of these approaches catches everything on its own, and detection will matter most when it is paired with clear labels and skepticism about content that arrives without a source.

Reporting based on an external source.