Is That Campaign Video AI? Google's Free Checker Can Now Tell You. Sometimes

A video of a candidate saying something outrageous lands in your family WhatsApp group. It looks real. It sounds real. Do you forward it?

Google has just made one part of that decision easier. Its SynthID Detector, a tool that checks photos, videos and audio for hidden AI watermarks, is now open to anyone in the world. Until now, only journalists and media professionals could test it. With Nigeria heading into the 2027 general elections, the timing matters.

What happened

Google DeepMind announced on October 7 that the detector is now available globally, in English. You sign in with a Google, Apple or OpenAI account, upload a file, and the tool tells you whether it finds a SynthID watermark.

SynthID is an invisible mark that some AI tools stamp into the content they create. You cannot see or hear it. Google says it survives common tricks like cropping, filters and compression for images and video, and added noise, speed changes and MP3 compression for audio.

The detector does not only catch Google's own content. It also reads watermarks from partners including OpenAI, NVIDIA and Kakao, and Apple is expected to join later.

The key numbers

180 billion+ images and videos watermarked with SynthID since it launched in 2023, according to Google

240,000 years worth of audio also carries the watermark

1 million+ verification requests a day already handled by checks built into Google Search, Gemini and Chrome

3 ways to sign in: Google, Apple or OpenAI accounts

The big catch

This is not a universal fake detector. It only finds content made with tools that use SynthID. A deepfake made with an app that does not add the watermark will simply come back clean.

So a "no watermark found" result does not prove a video is real or human-made. The tool also cannot tell you whether a genuine clip has been given a false caption, date or location, which is one of the most common tricks in election season.

What to do before you forward that clip

Run suspicious photos, videos or voice notes through the detector. A positive result is strong evidence the content came from a supported AI tool.

If nothing is found, keep checking. Look for the original source and whether credible news outlets have reported it.

Use a reverse image search on key frames to see if the picture or video appeared earlier in a different context.

Be extra wary of voice notes. Cloned voices are cheap to make and easy to spread on WhatsApp.

When in doubt, do not forward. One pause can stop a fake from reaching hundreds of people.

What this means for Nigerians

Campaign season will bring a flood of AI-made clips of candidates, fake endorsements and doctored audio, much of it moving through WhatsApp and TikTok rather than news sites. Voters, journalists, fact-checkers and party media teams now have a free public tool that can confirm when a clip came from Google, OpenAI or another supported AI generator. But because the detector is English-only for now and only catches watermarked content, it should be one step in your checking routine, not the final word. Content creators and agencies using Gemini or OpenAI tools should also expect their AI visuals to be identifiable, so label them honestly.

What to watch next

Watch for Apple to switch on its promised support, which would widen the net further. Also watch whether more AI companies adopt SynthID, and whether Nigerian fact-checkers and election observers build it into their monitoring ahead of 2027.

Sources: BusinessDay, Techpoint Africa, MacRumors