AI search engines have high error rates, study finds

A new study from the Columbia Journalism Review reveals that AI search engines, including OpenAI's ChatGPT and Google's Gemini, provide incorrect answers over 60% of the time. The most accurate model, Perplexity, still had a 37% error rate. The study tested eight AI models by asking them to identify details from articles. It found that many AI tools confidently gave wrong answers and often failed to cite sources correctly. For example, ChatGPT linked to the wrong source nearly 40% of the time. These findings highlight significant issues with the reliability of AI search engines compared to traditional search methods. The study raises concerns about the impact of AI on the quality of information and the online media economy.