Imagine scrolling through your feed and suddenly stopping at a photograph that makes your stomach drop. There, on a street in France, stands an elderly couple—both in their eighties—looking dazed and vulnerable, seemingly stripped of their clothes and robbed during a night of violent unrest. The image is designed to provoke immediate shock, outrage, and sympathy. It is shared with a warning: another brutal incident, another sign that the world is falling apart. Within hours, millions of people have seen it. They comment, they share, they argue. Some express fury at the government, others vent their anger at immigrants, and a few quietly doubt whether the photo is real. That doubt matters, because the images are not real at all. They were generated by artificial intelligence, then carefully spread online to look like documentary evidence. The story they tell—of an elderly couple being humiliated and robbed during riots in France—never happened. Yet because the images were so emotionally potent, they traveled across the internet faster than any correction could follow.
The false images found a particularly eager audience among far-right accounts on X, where they were amplified with inflammatory captions and apocalyptic warnings. Some accounts called the images “the most horrific image of October 17th,” while others declared that “France has fallen” and warned that the United Kingdom and the rest of Europe could be next. The alleged victims were described as Jewish, and French President Emmanuel Macron was accused of failing “the Israeli people.” Some went even further, comparing the fictional assault to a 1941 pogrom in Ukraine—a deeply painful historical reference used here to manufacture a sense of contemporary emergency. The American conspiracy theorist Alex Jones joined in, sharing one of the images with a long and hateful message that blamed the supposed violence on a “criminal consortium of Muslims and communists, who are unified by their hatred of the West and white people.” In his post, he claimed that an eighty-plus-year-old couple had been robbed in Paris and that “in the future, your lives will be taken in mass for the crime of being white.” It was a deliberately inflammatory narrative, built on a foundation of pixels that had no connection to reality. But the anger it generated was real, and it spread through the ecosystem of outrage politics exactly as designed.
So how do we know the images were fake? Journalists at France 24 used SynthID, an invisible digital watermark detection tool developed to help identify AI-generated content. This tool allowed them to confirm that both photographs had been created using OpenAI tools. The verification process did not stop there. The journalists traced one of the images back to a montage published late on October 17th by Infos-Israel.news, a pro-Israel website that had apparently shared the fabricated images without adequate scrutiny. Once the images were examined closely, the signs of manipulation became obvious. The elderly woman appeared to have only three fingers on one hand, with a deformed thumb. A bystander in the background seemed to have two heads. A fallen bollard in the street did not match the kind typically found in Paris. These are the kinds of mistakes that AI image generators still make, especially when creating complex scenes with human bodies and urban furniture. But in the heat of a breaking news event, few people pause to count fingers or inspect bollards. The image feels real because it mimics the visual language of news photography: grainy, chaotic, and immediate. The technology used to create it is advancing so quickly that even these flaws are becoming less common, which makes the challenge of verifying content more urgent than ever.
While the fabricated images were dominating discussions online, the real protests in France were being pushed aside. The demonstrations began in September, when students at a high school southeast of Paris protested against staff shortages and deteriorating school conditions. What started as a demand for better education soon grew into a broader movement, fueled by the cost-of-living crisis, rising fuel prices, and declining public health services. Across several major cities, the protests turned violent. High-street windows were smashed, shops were looted, and products stolen from stores later appeared for resale on platforms such as Vinted. In other words, there was a real story unfolding, one that involved genuine grievances and live social tensions. But instead of reporting on the reasons behind the protests, some outlets chose to sexualize and weaponize an invented image. Israeli right-wing news channel C14, for example, covered the French riots under the headline “Islam Conquers Europe,” and shared footage of youth gangs smashing windows in the streets. The footage was from Milan and other cities, not France, but that did not stop it from being used as evidence of a civilizational collapse. The truth was more nuanced and more human: students worried about their education, families struggling to pay for heating and food, nurses and doctors exhausted by years of underfunding. These are not abstract threats to a way of life; they are the material concerns of real people.
It is easy to dismiss AI-generated misinformation as a technical problem, a glitch in the information ecosystem. But that misses the harm it does to real people. The elderly couple in the images never existed, but the fear and hatred they inspired were felt by actual communities. Jewish families in France saw images of a supposed pogrom and felt the weight of history pressing down on them. Muslim citizens were blamed for a crime that no one committed. The public was taught to distrust not only the media but also their own neighbors. This is the real danger of AI-generated content: it does not just deceive individuals, it poisons collective trust. When a false image is shared widely enough, it becomes a reference point, a fake memory that shapes how people interpret real events. Later, even when the truth emerges, it often cannot undo the emotional damage. The image has already settled in people’s minds. They remember the outrage more vividly than the correction. This is why the verification work done by journalists matters so much, and why platforms like X must take greater responsibility for what they allow to spread. A fleeting label of “this image is AI-generated” is not enough when the post has already reached millions of people and triggered a wave of hostility. The algorithm that promoted the false image will not apologize. The accounts that posted it will not take it down. The damage simply becomes part of the background noise.
So what can be done? The first step is to acknowledge that the problem is not just technological, it is deeply human. We share what shocks us, what frightens us, and what confirms our existing beliefs. AI-generated images exploit these instincts with terrifying precision. Tools like SynthID are useful, but they are only one piece of the larger solution. Platforms need to commit to transparency, fact-checking, and rapid response when viral falsehoods are identified. Journalists need to keep doing the slow, meticulous work of tracing images and verifying sources. And ordinary internet users need to build new habits: hesitate before reposting, ask who created the image, look for obvious flaws, and trust reliable news organizations over anonymous accounts. None of this is as effortless as clicking “share,” but it is necessary if we want to live in a world where truth is still possible. The real story from France is not about an elderly couple stripped of their clothes. It is about a society struggling with real problems, about students and families demanding better, and about the people who tried to exploit that struggle by spreading lies. The next time you see an image that makes your blood boil, take a breath. Count the fingers. Look for the source. Ask why it was made. Because the fakes will only become more convincing, and our ability to see through them will only matter more.