I’m a Professional Fact-Checker. AI Is Wrong More Often Than You Think

Editor's take

A professional fact-checker found AI models like ChatGPT and Gemini to be unreliable for verifying information, frequently introducing errors or fabricating details. This challenges the notion of AI as an immediate solution for combating misinformation.

The implications are significant for media organizations and researchers who might otherwise offload verification tasks to AI. It highlights the current limitations of large language models in discerning truth from falsehood, a critical capability for maintaining information integrity in an increasingly digital world.

Future developments will likely focus on improving AI's reasoning and grounding capabilities, potentially through hybrid human-AI systems or more sophisticated training data. The key question remains whether AI can transcend its current probabilistic nature to achieve genuine factual accuracy, or if it will remain a tool requiring constant human oversight.