The images looked like evidence of a dramatic protest in Portsmouth. Crowds gathered in the street. Police surrounded demonstrators. Masked men stood in the sea carrying a nationalist flag.
There was just one problem: the pictures were fake.
Two images circulated on social media following the anti-migrant protests in Portsmouth on 6 September were generated or edited using artificial intelligence. The protests themselves were real, with hundreds of demonstrators gathering after around 140 migrants arrived by boat, but the particular images being shared were not.
The fakes contained several clues. Both carried a SynthID watermark associated with AI-generated or AI-edited content. Faces appeared blurred or distorted; one image showed Portsmouth’s Spinnaker Tower with an incorrect colour, while another contained gibberish writing on a sign. Neither image appeared to match the actual location of the protest.
Yet that did not stop people sharing them as if they were genuine.
And that is where the bigger story begins.
AI gives extremists a powerful new propaganda tool
Extremist movements have always used propaganda. What has changed is the speed, scale and sophistication with which it can now be produced.
Generative AI allows users to create convincing photographs, videos, audio and graphics from simple instructions. Researchers and organisations monitoring extremism have documented the technology being used to spread racist, antisemitic, conspiratorial and extremist material. The Anti-Defamation League, for example, has found AI tools being exploited to create hateful music, propaganda and other extremist content.
Research into AI-generated hate has also found systematic use of synthetic imagery to portray minorities as dangerous, criminal or threatening. A 2025 study of more than 1,300 AI-generated images and videos found recurring themes including dehumanisation, conspiracy theories, sexualisation and the aestheticisation of violence.
The technology therefore does not simply make existing propaganda easier to produce. It makes it possible to manufacture entirely new “evidence”.
An extremist account can claim that migrants have attacked a town, that police have lost control, or that a huge nationalist mobilisation has taken place — and then create an image apparently proving the claim.
The photograph never needed to exist.
From misinformation to mobilisation
This matters because propaganda works through emotion.
A fabricated image of a violent migrant, a burning building, or an enormous protest can generate fear and anger before anyone has time to check it. Once shared repeatedly within communities that already believe the underlying narrative, the distinction between fact and fiction can disappear.
Full Fact has warned that AI-generated imagery shared as genuine can gradually undermine trust in information online. It has also documented manipulated and AI-generated material being used alongside false narratives about immigration, Muslims and social unrest in the UK.
The danger is particularly acute around events such as protests, terrorist attacks, elections and incidents involving immigration. These are moments when people are looking for information quickly and emotions are already running high.
Extremist propagandists understand that.
The Portsmouth images demonstrate the tactic perfectly. There was a real event, real protesters and genuine footage. AI-generated material could therefore be mixed into an existing news story, making the fabricated images appear much more credible.
The fake does not have to replace reality. It only has to distort it.
The new challenge for social media
AI also creates a problem for platforms and fact-checkers. A fabricated image can be produced in seconds and potentially reach thousands of people before it can be investigated.
And increasingly sophisticated AI means that the old advice about looking for obvious mistakes, strange hands, distorted faces, or nonsensical text—will become less reliable.
The answer cannot simply be to distrust every photograph. Nor should AI itself be regarded as inherently dangerous. The technology has enormous legitimate applications.
But we do need to change how we consume information.
Before sharing a dramatic image, ask who originally posted it, whether reputable news organisations have published independent photographs or video, whether the location matches what is being claimed and whether other sources confirm the event.
Most importantly, remember that an image is no longer necessarily evidence that something happened.
For extremists and racists, AI offers the ability to manufacture propaganda at unprecedented speed. The objective is not necessarily to persuade everyone. Sometimes it is enough to confuse people, reinforce existing prejudices and make a false narrative feel possible.
The Portsmouth images are therefore more than two examples of fake photographs. They are a glimpse of a future in which right-wing extremist propaganda can be manufactured almost instantly and in which the battle over truth is increasingly being fought not only with words but also with synthetic realities.
