ChatGPT did not cure a dog's cancer: how a viral story turned out to be exaggerated

The story of how ChatGPT cured a dog of cancer spread across the internet at a speed that Big Tech could only welcome — living proof that artificial intelligence is ready to revolutionize medicine and defeat one of the deadliest diseases. Reality, as usual, turned out to be far more complex. In 2024, Sydney tech entrepreneur Paul Conyngham learned that his Staffordshire Bull Terrier, Rosie, had cancer.

Chemotherapy slowed the disease's progression, but the tumors did not shrink. When veterinarians said "nothing can be done," Conyngham — a man without professional education in biology or medicine — turned to ChatGPT. According to him, with the help of the neural network, he managed to develop something like a personalized treatment plan.

The dog survived, and her owner attributed this to AI's merit. The story first appeared in Australian publication The Australian and immediately began to spread: it was picked up by technology media, bloggers, and AI enthusiasts around the world. The narrative had everything needed for viral material — a touching hero, a deadly disease, a hopeless situation, and rescue through technology.

This is exactly the kind of case that proponents of AI application in medicine have long been waiting for. However, journalists at The Verge, examining the details, discovered that the picture is significantly more complex. ChatGPT's role in Rosie's recovery cannot be independently verified.

It is unknown what exact prompts the chatbot provided, how medically accurate they were, and most importantly — what exactly influenced the outcome. Oncological diseases in animals, like in humans, can have unpredictable courses, and "nothing can be done" from a veterinarian's lips does not always mean a final verdict. This story is a good example of how the media environment around AI functions as an amplifier.

An anecdotal case, presented from the right angle, becomes "proof" of a revolution. At the same time, questions about methodology, reproducibility, and actual cause-and-effect relationships do not make headlines. The important conclusion here is not whether ChatGPT is good or bad as a medical tool — that is a separate and serious discussion.

It is important that individual success stories, especially emotionally saturated ones, are poorly suited for evaluating the real capabilities of the technology. AI is indeed actively being integrated into medical research and diagnostics — with real results and peer-reviewed data. But these stories are told much more quietly than the story of the rescued dog.