A scientific prank has laid bare the dangerous gullibility of artificial intelligence models. In 2024, Almira Osmanovic Thunström, a doctoral candidate in medical technology at the University of Gothenburg, uploaded two phoney studies on a fake eye condition called bixonimania. Soon enough, AI chatbots were waxing eloquent on it and advising people what to do if symptoms surfaced. She did it to show how AI systems trawl the internet to gather ‘knowledge’ and can be coaxed to blur the line between the truth and a lie. This underlines the grave risk that Indians, who are increasingly treating chatbots as experts to understand lab reports and offer medical advice, are putting themselves in. What not to do with AI should be part of the school curriculums trying to evangelise the technology.