Théodore Cazals was scrolling through Facebook one evening when he came across an image that stopped him cold. On a rain-soaked road, two young children sat beside a birthday cake. The rain was coming down hard, and the children looked emaciated, their cheekbones sharp against their skin. A handwritten sign nearby asked viewers to like the photograph because it was supposedly the child’s birthday. The post seemed designed to make people feel something: pity, tenderness, a desire to acknowledge a moment of joy in the midst of hardship. And it worked. Almost one million users clicked like or added a heart emoji. But Cazals, a 20-year-old student from Paris, could not bring himself to click. There was something profoundly wrong with the picture. The children, though dressed as young boys, had thick, unruly beards covering their faces. One of them appeared to have no hands and only a single foot. They were sitting cross-legged in the middle of a busy road in a torrential downpour, as if the rain did not bother them at all. The more he looked, the more it became obvious: this was not a photograph. It was an image generated by artificial intelligence, a cheap and strange construction that blended real human suffering with impossible details. And yet it had racked up nearly a million reactions. He was stunned, not just by the picture itself, but by what it said about the online world. People were willing to believe almost anything if it made them feel something. The absurdity did not matter. The lies did not matter. The only thing that mattered was the emotional pulse, the twitch of sympathy that led to a like.
Cazals did not have a plan at first. He was just angry and a little amused. He decided to start an account on X under the name “Insane AI Slop” and began posting examples of fake, unconvincing images and videos that were spreading across social media. His tone was mocking, but his purpose was serious. He wanted to expose the strange world of AI-generated content, the oddly proportioned hands, the melting faces, the impossible backgrounds, the emotional manipulation that hid behind a veneer of heartwarming stories. He did not have to wait long for material. Within days, people began sending him their own examples. The account grew quickly, from a handful of followers to tens of thousands, and eventually beyond 133,000. It had become a kind of community, a group of digital detectives who traded the strangest examples of AI slop they could find and tried to understand why these images kept appearing. The BBC, which later profiled Cazals and his campaign, described it as part of a wider backlash against low-quality AI content. The term “slop” was key. It suggested not just bad art, but a careless, industrialized process, content produced quickly and cheaply to be consumed and forgotten. Unlike a deepfake, which might be crafted to deceive in a sophisticated way, AI slop often did not even try to be convincing. It relied instead on emotional triggers, on images of hungry children, brave soldiers, angels descending from heaven, or cats doing impossible things. The goal was not to create beauty or truth but to generate engagement. And in that, it was remarkably successful.
As the submissions flooded in, Cazals began to notice patterns. Certain themes appeared over and over, as if the AI programs had been trained on the same sentimental clichés. Religion was one of them: majestic images of Jesus with glowing eyes, angels watching over families, crosses in dramatic skies. Military subjects were another: soldiers in heroic poses, often depicted with an exaggerated sense of sacrifice and patriotism. But the most common theme, and the one that disturbed Cazals the most, involved poor children. In image after image, children from Africa, Asia, or Latin America were shown performing remarkable feats of kindness or creativity. A child might be sculpting an incredible statue out of discarded trash. A teenager might be carrying an elderly person through a flood. A young girl might be sharing her last piece of bread with a stray dog. These images were designed to inspire awe and warmth, and they were wildly popular. “Kids in the third world doing impressive stuff is always popular,” Cazals observed, “like a poor kid in Africa making an insane statue out of trash. I think people find it wholesome so the creators think, ‘Great, let’s make more of this stuff up.’” This was the heart of the problem. The AI was not simply inventing unrealistic scenes; it was exploiting a long tradition of paternalistic storytelling about poverty. The images reduced complex human lives to simple parables of resilience and gratitude. They made poverty look noble and suffering look inspiring. They also made it harder for real stories, real photographs, and real people to compete for attention. Why would a platform show you an actual documentary about a community in need when an AI-generated picture of a smiling, bearded child can deliver the same emotional reward with none of the complexity?
Cazals soon realized that the problem extended far beyond Facebook and X. He began to notice strange AI-generated cartoons on YouTube, videos that had millions of views and seemed to target very young children. Some were just bizarre, combining nursery-rhyme aesthetics with nonsense plots. Others were darker. One video showed a cat story involving deadly belly parasites. Another showed a woman who ate a parasite, turned into a giant monster, and was eventually healed by Jesus. The animations were crude, but they had a hallucinatory quality that might easily frighten a small child. Cazals considered them not just disturbing but harmful, and he decided to use his growing influence to do something about it. He contacted YouTube moderators and drew attention to the offending channels. To his surprise, his alerts led to action. YouTube removed the channels, saying they had violated its community guidelines. The company also released a statement about its broader approach to AI content, saying it aimed to connect users with high-quality content regardless of how it was made and that it was working to reduce the spread of low-quality AI material. It was a small victory, and Cazals knew it. For every channel removed, new ones seemed to appear overnight. The flood did not stop. But the episode showed that platforms could act when they chose to, and that a single person with enough persistence could force a conversation. It also highlighted a deeper anxiety: if AI-generated content was already this strange and uncontrollable on mainstream platforms, what would happen as the technology improved and the images became increasingly indistinguishable from reality?
The backlash that Cazals helped inspire did not stay confined to his account. Across the internet, on TikTok, Threads, Instagram, and even in the comments of Facebook posts, users had started calling out AI-generated content. The BBC report noted a remarkable shift: in many cases, comments criticizing the AI nature of a post received more likes than the original post itself. This was a sign that the audience was wising up, at least a little. It was no longer embarrassing to point out the obvious. In fact, it was almost a sport. But the response was not uniform. Emily Thorson, a professor at Syracuse University, explained why the reaction to AI slop depended so much on context. “If a person is on a short-video platform solely for entertainment, then their standard for whether something is worthwhile is simply ‘is it entertaining?’” she said. “But if someone is on the platform to learn about a topic or to connect with community members, then they might perceive AI-generated content as more problematic.” This insight is crucial. People who use social media for fun may not care if an image is fake, as long as it is amusing or moving. A video of a cat with an extra set of paws is just a joke. But for someone searching for information about a health condition, a political event, or a community issue, AI slop is a serious hazard. It undermines trust, wastes time, and can lead people to make decisions based on falsehoods. The platforms, of course, have little incentive to solve the problem. AI slop is cheap to produce and highly engaging. It keeps users on the site longer, which increases ad revenue. The cost is externalized, borne by users who have to wade through an ever-growing sea of nonsense. The question is how long that bargain will be accepted.
For Cazals, the answer is clear: the bargain must be rejected. He did not consider himself an enemy of artificial intelligence. “Unlike a lot of my followers, I’m not dogmatically against AI,” he said. “I’m against the pollution online of AI slop that’s made for quick entertainment and views.” It was a distinction that set him apart from more radical critics. He was not calling for a ban on AI or a return to some pre-digital past. He was calling for accountability. He wanted platforms to label AI-generated content, to remove content that was deceptive or harmful, and to stop rewarding the factories that churned out fake images by the dozen. He also wanted viewers to develop a kind of reflex, to pause before clicking like on a sentimental image and ask whether it might be a lie. His campaign was a reminder that the internet is not just a place of passive consumption. It is also a place where individuals can organize, investigate, and push back. The community he built around “Insane AI Slop” was small compared to the billions of users on social media, but it demonstrated that not everyone is willing to accept the degradation of public truth. The problem will only grow as AI becomes more sophisticated. The images that once looked absurd will soon look flawless. The bearded children will disappear, replaced by perfect fakes that no human eye can detect. But the underlying issue will remain the same: our attention is the most valuable commodity on earth, and there are countless actors, human and machine, trying to steal it. Cazals’ story is not a cure. It is a warning, and a small, stubborn act of resistance.