Full Fact says video of thousands at a London protest is AI-generated

A Facebook video that appears to show thousands of people protesting outside the Houses of Parliament is AI-generated, according to a fact check published by Full Fact on 25 September 2026.

What the video shows

The UK fact-checking charity says the post is captioned "Wake up, England" and "seemingly shows a large crowd of people in central London, many of whom are carrying Union Jack flags". Full Fact says it has rated the claim as altered, under its work fact checking potentially false pictures, videos and stories on Facebook, "because this video isn't real and has been made with artificial intelligence".

The fact check does not say who made the video or who first shared it.

How Full Fact established it was AI

The fact check sets out two pieces of evidence:

• A hidden watermark. "The video contains an invisible SynthID watermark that shows it was made using Google's AI tools."

• Visual glitches. "A number of visual glitches are present in the video, including garbled text on a road marking, and buildings and structures like Westminster Bridge that do not match their positions in real life."

Full Fact's own guide to spotting AI content, published in April 2025, explains how it has used that kind of watermark before. It says Google had told it that content produced with its AI products "contains an embedded 'SynthID' digital watermark that is not visible to the human eye". The guide adds that Google also told Full Fact "the presence of a watermark can't tell us whether AI was used to completely generate a brand-new image or just modify an existing one".

How to check videos like this, according to Full Fact

The fact check points readers to Full Fact's guide to spotting AI content and its misinformation toolkit, which it says "can help verify what you're looking at before you share it with others". The guide describes the methods Full Fact relies on as "analysing the media itself, using reverse image search tools, asking whether a given scenario is plausible, and looking for inconsistencies". Among the points it makes:

• Reverse image search can reveal hidden watermarks. In one earlier case, Full Fact says, putting an image through Google's reverse image function showed, under the "about" section, that it was flagged as "Made with Google AI".

• Visible watermarks are not always there. The guide says they "can easily be cropped away before an image is shared", and that content from a paid AI tool can often be generated without one.

• Look for inconsistencies. The guide says fingers, arms, ears, toes and teeth "are often parts of the body where errors may appear", and that "at first glance you might miss these discrepancies".

• Slow down. "Before sharing striking images like this you find online, we advise taking a second to look even closer."

• Be wary of detection tools. Full Fact says online tools that claim to verify AI content have not proved reliable in its experience, and that it does not routinely use them.

The limits Full Fact acknowledges

The guide says that "identifying these kinds of images and videos as fakes is becoming increasingly difficult", and that Full Fact often has to caveat its assessments by saying content is "likely" or "almost certainly" AI-generated. The London fact check carries no such caveat: it says the video "is AI-generated".

On our reading, this case combines two of the kinds of evidence the guide describes: a hidden watermark, and inconsistencies in the scene itself.