AI Wildlife: Bogus Stories and Big Lies in Bighorn Country

Video makes this considerably more complicated. During my investigation, I found multiple AI-generated videos involving mountain goats and other mountain ungulates. Some showed animals negotiating cliffs or performing maneuvers that stretched credibility even for animals famous for navigating extreme country.

One looked remarkably convincing until an ibex suddenly appeared among what were supposed to be North American mountain goats. Another supposedly showed action-camera footage from a mountain goat traveling a narrow ledge and encountering a snow leopard. Spectacular footage, except it was geographically impossible.

North American mountain goats don't share their wild range with ibex or snow leopards. The AI understood the visual ingredients of horned mountain animals, sheer cliffs and spectacular predators, but it didn't understand natural history.

Another video concerned me for a different reason. It showed a man standing face-to-face with an apparently unrestrained ibex while casually attaching a GoPro-style camera to the animal. The ibex stood there as if putting a camera on a wild mountain ungulate were about as complicated as putting a collar on the family dog.

That's not how wildlife research works. Capturing, handling, collaring and monitoring wild mountain ungulates requires planning, permits, specialized equipment, trained professionals, funding and considerable effort. AI can erase all of that in seconds and give viewers “wildlife research” that required none of it because the event never happened.

Kendrick Chittock, operations director for the Rocky Mountain Goat Alliance, said checking the source is increasingly important as synthetic wildlife content becomes more common.

“AI-generated wildlife content is becoming more prevalent. The best way to tell if something is real or accurate is to check the source. Agencies and conservation organizations are the most reliable places for wildlife conservation information, but anecdotal videos and images from real people in the mountains are also just as entertaining and do not offer inaccurate species overlap or exaggerated feats,” Chittock said. 

“Wildlife is amazing in its own right. We don't need AI to enhance it.”

Corey Mason, chief operating officer and executive vice president of conservation for the Wild Sheep Foundation (WSF) and a Certified Wildlife Biologist, has watched the verification challenge change dramatically during his career.  Before joining WSF, Mason spent 16 years with the Texas Parks and Wildlife Department, serving as a management area biologist, program leader and regional director.

“AI has created an environment where nothing can be trusted. When I worked for a state game and fish agency, we often went to the field to verify game camera pics of large predators in marginal and new areas to ensure they were not altered and that the site was where they claimed it to be. Now, images can be completely created using the background you provide, so it makes truthing images very challenging,” Mason said.

Think about how much has changed. Wildlife professionals have always dealt with questionable photographs, unusual sightings and claims requiring verification. There was a time when checking an extraordinary game-camera image might mean going into the field and examining the terrain, vegetation and other physical evidence to determine whether the photograph was actually taken where someone claimed.

Now the animal and the landscape can both be manufactured.

As I dug further into these AI-generated wildlife stories, I came across research showing my concerns went well beyond what I was finding on social media.

José Guerrero-Casado and four colleagues examined the issue in a peer-reviewed paper, “Threats to Conservation From Artificial-Intelligence-Generated Wildlife Images and Videos,” published in Conservation Biology.

“AI-generated fake images and videos can distort people's perceptions of wildlife and lead to the spread of misinformation,” the researchers wrote.

That was exactly what I was seeing.

The researchers examined unrealistic relationships between species, exaggerated animal behavior and other fictional representations that can distort people's understanding of wildlife. They also addressed the ability to produce this material rapidly and at scale and warned about audiences that may lack the wildlife knowledge necessary to recognize what isn't real.

That last point particularly concerns me. A generation of young people is growing up with social media as one of their primary windows into wildlife. Someone who has never seen a mountain goat in the wild may have no reason to question an incredible AI video. They may not know an ibex doesn't belong in the scene or that a snow leopard doesn't occur anywhere near a North American mountain goat.

We're potentially creating a false natural history in real time.

There is another part of this the wildlife conservation community needs to consider. The real stories are already incredible.

WSF, its chapters and affiliates, and state, tribal and provincial wildlife agencies are doing legitimate conservation work involving habitat, disease, restoration, translocations, research and wildlife management. That takes money, expertise, cooperation and often years of commitment.

RMGA is doing the same kind of real-world work for mountain goats. Real people are putting boots on mountains and resources into conservation as synthetic wildlife competes for the public's attention.

A sensational AI-generated video can attract far more attention than a legitimate sheep translocation, goat count or research project that actually contributes to conservation. Wildlife professionals can spend years studying an animal while a computer manufactures a more spectacular version of that animal in seconds.

The implications go beyond entertainment. The same technology being used to create fake wildlife encounters could be used to influence public opinion about hunting and wildlife management. Animal-rights or anti-hunting interests could manufacture emotionally powerful imagery or narratives designed to influence attitudes about hunters and hunting.

I asked Mason whether he could envision animal-rights interests using AI-generated words or imagery to appeal to emotions and undermine science-based wildlife management.

“With what I have witnessed over the last couple of decades, I can absolutely see that, especially to push the desired narrative into the news cycle, true or not,” Mason said.

That possibility becomes particularly concerning when wildlife issues reach the ballot box.

Imagine a realistic-looking video appearing days or weeks before a wildlife-related ballot measure. It shows a wounded animal suffering after a hunt, a terrified young animal supposedly separated from its mother or an animal struggling while a hunter stands nearby. The animals look real. The habitat looks real. The people look real.

But none of it happened.

By the time wildlife professionals, conservation organizations or journalists determine the footage is artificial, millions of people could have already seen it. Some may have shared it specifically because it made them angry, sad or horrified. The emotional impact occurs before the verification process ever begins.

That is where AI-generated wildlife content could become something far more consequential than clickbait.

Wildlife issues have always had an emotional component. Images of injured animals can be powerful. Images involving young animals can be even more so. Now those images don't necessarily have to document something that actually happened. The animal can be generated. The injury can be generated. The hunter can be generated. The entire event can be generated. And it can all be packaged to look like evidence.

That changes the equation. A campaign built around an actual photograph or video can be challenged by examining where it was taken, who recorded it and what happened before and after the camera was rolling. Artificial intelligence introduces the possibility of something different: an emotionally devastating piece of supposed evidence for which there was never an original event to investigate.

That doesn't mean a particular organization is doing it, and I'm not making that accusation without evidence. The point is that the capability exists, and hunting is particularly vulnerable because images involving animals can generate immediate and powerful emotional responses.

A photograph once required someone to be there. Video required a camera pointed at an actual event. Neither of those assumptions can safely be made anymore.

Hunter-conservationists need to understand what we're dealing with. Before sharing the wildlife story that makes your blood boil, or the one that's simply too amazing not to share, take another minute. Who reported it? Is a wildlife agency identified? Is a biologist named? Can the authentic research be found? Does the species actually occur where the video supposedly was filmed?

And keep reading. Sometimes the little disclaimer at the bottom changes everything you just saw.

Artificial intelligence itself isn't necessarily the enemy. It can be a useful tool, including in wildlife science and conservation. The problem is an information environment in which fiction can be manufactured faster than truth can be verified.

Wild sheep restoration doesn't need to be fictionalized to be fascinating. The real work being done by hunter-conservationists, WSF and its chapters and affiliates, and state, tribal and provincial wildlife professionals deserves to be seen. Unfortunately, some of that conservation work is now competing for attention with wildlife stories that never happened and animals doing things they never did.

An AI-generated wildlife story can be created in seconds. Real conservation takes time. Restoring habitat, rebuilding wildlife populations, conducting research and putting animals back on landscapes can require years or even decades of work by people committed to getting it right. But there is a potentially darker side of the artificial wild that deserves its own examination.

What happens when the goal isn't simply getting a click? What happens when AI-generated wildlife content is designed to create outrage, change public attitudes about hunting or influence a larger wildlife debate? And what happens when the artificial part isn't limited to the photograph or video?

AI-generated influencers and increasingly capable AI agents raise another question. A synthetic persona can generate posts and comments, interact with people online and produce content at a scale no individual person could match. These systems still require human-created goals, instructions or other direction, but they increasingly can carry out parts of that work with less human involvement.

That opens another potential avenue for manipulation. An AI-generated wildlife post could theoretically be amplified by accounts producing comments, arguments and reactions that also aren't coming from real people. Someone encountering the discussion might see what appears to be a groundswell of outrage without realizing that some of the voices helping create that impression could be artificial.

Take that one step further. What happens when an AI influencer is given an anti-hunting objective and allowed to continually generate content, interact with followers and respond to developing wildlife issues? The person behind it might not need to write every post, create every image or personally make every argument.

The wildlife might be artificial. The influencer might be artificial. Even some of the apparent public reaction could be artificial. That is where this investigation goes next.

In Part Two, we'll dig deeper into this side of the artificial wild, including the potential for AI-generated wildlife imagery and narratives to influence attitudes about hunting and science-based wildlife management. We'll examine AI influencers, synthetic comments and other forms of manipulation, along with how increasingly convincing and increasingly autonomous AI content could shape the wildlife debate and what hunter-conservationists need to watch for as this technology continues to evolve.

The artificial wild is easy to manufacture. The real wild is challenging to conserve.

Contributing Author: Chester Moore is an award-winning wildlife journalist, wildlife photographer, and lifelong hunter from Texas. He operates the Higher Calling Wildlife® blog and podcast and contributes to many outdoors publications.