By Jack Keller

A two-headed black bear saunters down a wooded road and video of it is seen by 4.3 million X viewers.  Dozens of commenters pester Grok, X’s AI chatbot, “@Grok, is this real?”

“The video is real footage, but it’s not a two-headed bear. It’s an optical illusion: a mother black bear carrying her cub on her back, with the heads aligned to look like one body with two heads,” Grok reassures. “Bear cubs often ride like this for safety. While the illusion is convincing, no verified cases of adult two-headed bears exist in wildlife records.”

The pedantic explanation is wrong. The clip isn’t an optical illusion; it was produced by generative AI—software capable of creating images and videos that look real—and the platform’s built in AI fact-checker that millions of users rely on couldn’t recognize it.

Two headed bears aren’t real, despite being popular in folklore. And this video was hardly an outlier. Fabricated animal content now flows steadily through social media, often far more subtly than the two-headed bear.

A generated image of a “Mommy shoebill and her baby” racked up millions of views on X. The hyper realistic, yet stylized video portrayed their two faces appearing to angrily frown. “The baby came out of the egg already tired of everyone’s nonsense,” a commenter jokes. Only when the image is placed side by side with a real photo of a shoebill can a person see that the viral image is artificial.

In another video seen by 3 million viewers, a tabby kitten is sitting in a bubble bath. Bigger bubbles from around its backside are coming out and the kitten gives the camera a guilty look —the cat is passing gas.  It’s engineered to be funny and cute. But it’s also passing as real, while normalizing a situation most cats would find terrifying; it’s not difficult to imagine someone sticking their cat in a tub after watching that clip.

AI creates scenes that look real but never happened. By showing animals in unnatural situations, AI-generated content distorts how viewers understand wildlife behavior.

In a 2025 study published in Conservation Biology, researchers warned that AI-generated animal content is not just misleading, it could potentially reshape public understanding of animal behavior at a scale conservationists have never dealt with before.

Fake images can make endangered species seem more abundant, weakening people’s motivation to support life-saving initiatives. It can also increase demand for illegal pets, and inflame anxieties about wildlife encounters.

The study found three characteristics of social media that are exacerbating the issue: the ubiquity of social media, AI’s tendency to humanize animal behavior, and a growing, general disconnect from nature.

For most Americans, animals are no longer encountered in forests but instead on their feeds., Roughly 84% of U.S. adults use YouTube, 71% use Facebook, and half are on Instagram; platforms where animal videos and images routinely rack up millions of views, according to the Pew Research Center.

Generative AI feeds directly into this disconnect. Researchers in the Conservation Biology study noted it’s likely that AI-generated content will become more common than content produced by humans, meaning that fake portrayals of animals could soon set the standard for what viewers think animals look and behave like.

These fabricated images tend to anthropomorphize animals; it’s scripted content after all. One  recent example portrayed a curious squirrel investigating a suburban back deck, where homeowners put out ramen noodles and a glass of water for hungry wildlife. It’s designed to look like a hidden camera, deepening the deception. The squirrel is caught nibbling on the spicy noodles and standing on its hind legs for an animated, burning mouth reaction and going for the water.

Commenters expressed outrage, but for the wrong reasons—“this is ridiculous, exploiting the animal’s hunger for views,” and “if it was for him, why’d you put the pepper?”—even though the behavior and the clip is entirely fictional. This orchestrated scenario demonstrates how AI can manipulate understanding and empathy, reinforcing misperceptions about animals.

People are more susceptible to believe in fabricated animal content when they’re already severed from the natural world, the researchers noted.  For most Americans, that disconnection is palpable. Most adults, according to The Nature of Americans, now spend five or fewer hours a week in nature—and they’re satisfied with that low amount.

This disconnect from nature is making people more gullible, and that impressionability is putting real animals at risk.

In St. Louis in January, 2026, viral vervet monkey sightings led St. Louis into a frenzy. The St. Louis Department of Health elected to call off a search for four monkeys, who were thought to have escaped from a residence after a series of AI-generated images of monkeys roaming the city circulated online, muddying the waters between actual tips and AI-generated hysteria.

While the frenzy almost certainly started with a real sighting, AI experts couldn’t discern whether the viral images were real or not, and when it came time for the city to make choices about where to allocate their resources, they chose not to act based on potentially-fabricated information.

It’s very possible that four monkeys imprisoned as pets ——owning a primate as a pet is highly dangerous, inhumane, and not to mention illegal in St. Louis— missed out on their chance for freedom due to opportunistic posts and viral jokes, where the pursuit of clicks overshadowed any real concern for the imperiled primates.

AI’s defenders will claim that it can produce educational content without needing to encroach upon an animals’ habitat. But even if that were true, the reality is that the vast majority of AI animal content is designed for engagement, not education.

To make matters worse, AI-generated content comes with a massive hidden cost.

Beyond deceiving viewers and putting actual animals at risk, AI-generated animal content also carries a staggering environmental toll, pitting it directly against the work of real conservation efforts.

According to an investigation conducted by the MIT Technology Review, a single AI-generated, 5-second video produced on the least efficient model available requires roughly the equivalent electricity needed to ride 38 miles on an e-bike or run a microwave for an hour, just for one video. Training these models—the process that teaches them to fabricate animal scenes in the first place—requires exponentially more energy.  Multiply that cost by millions as these clips are produced at an industrial scale, solely in pursuit of engagement.

The irony is stark: the environment is being strained to generate images of animals who never existed, while conservationists struggle to protect the animals who do.