TL;DR
A wave of AI-generated deepfake videos is spreading across TikTok and Instagram, depicting fictional scenes of urban decline in UK cities — particularly Croydon in south London. Collectively, these videos have racked up millions of views and are fuelling racist narratives, despite many viewers believing the content is real.
The Croydon Deepfakes
The BBC tracked down the creator behind the trend, who uses the handle RadialB. His AI-generated videos show fictional “taxpayer-funded” facilities — water parks, zoos, aquariums — filled with young men in balaclavas and padded jackets. One video depicting “roadmen in Parliament” reached eight million views in a single day.
RadialB, who is in his 20s, from north-west England, and has never been to Croydon, says his content is meant to be funny. But he also wants viewers to believe the scenes are real: “If people saw it and they immediately knew it was fake, then they would just scroll.”
Dozens of copycat accounts have since emerged, with users from as far as Israel and Brazil sharing the videos. Several accounts that appear to be based in the Middle East have posted multiple videos about London’s supposed decline.
Beyond Satire
The AI deepfakes fit into an existing trend of “decline porn” — content that portrays Western cities as overrun with immigrants and crime. While some content uses real footage taken out of context, AI is increasingly being used to fabricate scenes entirely.
The impact is measurable. YouGov data from January shows that a majority of Britons now believe London is unsafe — though only a third of people surveyed in the capital agree, and 81% say their own local area is safe.
RadialB acknowledges the videos provoke political reactions, noting that “50-year-olds and 60-year-olds” rage in the comments with “all this political stuff.” His TikTok account was banned for graphic content, but he has since set up a new one sharing similar videos.
Looking Forward
The trend highlights how improved AI video tools have dramatically lowered the barrier to creating convincing disinformation. Platform labelling policies exist — some videos carry small “AI-generated” labels — but they clearly are not preventing viewers from treating the content as real. For UK policymakers, the question is whether existing platform rules are keeping pace with the speed and scale of AI-generated disinformation.