Anyone can now check if a picture was made by AI with Google's SynthID site

Google has opened a website that lets anyone check whether an image, video or audio clip was made using artificial intelligence.

The tool was first made available to selected journalists, media professionals and researchers for testing last year. Google has now opened it to everyone.

How it works

The site looks for SynthID, an invisible digital watermark that Google introduced in 2023. The watermark is added to media created by Google's AI models, including Nano Banana for images, Veo for video and Lyria for music, and by tools such as Gemini, Flow and Vids.

The watermark is designed to survive common changes such as cropping, compression and filters, and is not visible to the eye or audible to the ear.

OpenAI, Nvidia and the South Korean company Kakao also support SynthID, and Apple is said to be adding support soon. OpenAI also runs its own site for checking content.

What you can check

• Images: JPG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF and GIF files.
• Video: MP4, MOV and WEBM files.
• Audio: WAV, MP3, OGG, FLAC, AAC and M4A files.

Google has also built SynthID checks into the Gemini app and the Chrome browser. It says people already make around one million requests a day to verify content.

The limits

A check can only find a watermark that was put there in the first place. Microsoft and Meta use their own standards for marking and verifying AI content, and many AI tools, including some open-source models, add no watermark at all. That means a clean result does not prove that a picture or clip is genuine.

Detection tools in general are not infallible, and often fail to identify content made by rival companies' models.

Why it matters

AI-generated images, video and voices have become very hard to tell apart from the real thing. They have been used in scams, in fake celebrity endorsements, in political disinformation and in harassment, including fake explicit images of real people.

Tools like SynthID are part of a wider push for provenance: ways of recording where a piece of media came from and how it has been changed. Supporters say this will become as routine as checking a website's padlock icon.

How to spot a fake

Alongside tools like this, fact-checkers suggest some simple steps before sharing a dramatic picture or video:

• Check who posted it first, and whether trusted news organisations are reporting the same thing.
• Look closely at hands, text, reflections and backgrounds, where AI images still often go wrong.
• Use a reverse image search to see whether the picture has appeared before in a different context.
• Be especially wary of anything that asks for money, or urges you to act quickly.

If in doubt, it is safest not to share it.