Artificial intelligence is ineffective and potentially harmful for fact checking

Fact checking can be an effective strategy against misinformation 1,2,3, but its implementation at scale is impeded by the overwhelming volume of information online 4. Recent artificial intelligence (AI) language models have shown impressive ability in fact-checking tasks 5,6, but how humans interact with fact-checking information provided by these models is unclear. Here we investigate the impact of fact checks generated by a popular AI model on belief in, and sharing intent of, political news in a preregistered randomized control experiment. Although the AI performs reasonably well in debunking false headlines, we find that it does not significantly affect participants’ ability to discern headline accuracy or share accurate news. However, the AI fact-checker is harmful in specific cases: it decreases beliefs in true headlines that it mislabels as false and increases beliefs for false headlines that it is unsure about. On the positive side, the AI increases sharing intents for correctly labeled true headlines. When participants are given the option to view AI fact checks and choose to do so, they are significantly more likely to share both true and false news but only more likely to believe false news.

Introduction. Digital misinformation has rapidly become a critical issue of modern society7,8. Recent work suggests that misinformation can erode support for climate change9,10, contribute to vaccine hesitancy11,12,13, exacerbate political polarization14, and even undermine democracy15. As a mitigation strategy, fact checking has proved effective at reducing people’s belief in3,1,16 and intention to share17 misinformation in various cultural settings2. However, this approach is not scalable, greatly limiting its applications4. To tackle this challenge, researchers and social media platforms have been exploring automated methods18 to detect misinformation19,20 and fact-check claims21,22,18,23,24,25. A robust fact-checking system must possess the ability to detect claims, retrieve relevant evidence, assess the veracity of each claim, and yield justifications for the provided conclusions26,27. Previous work attempting to meet these goals typically adopts cutting-edge artificial intelligence (AI) methods, specifically natural language processing.

Discussion / Conclusion. While our experimental design allows us to assess the causal effects of AI fact checking on the discernment of true and false news, it is important to exercise caution when generalizing the current results to different contexts. First, we use a specific version of ChatGPT as an automated fact-checking system; the findings may not be directly applicable to other fact-checking AI models. Second, the survey may not fully capture the complexities of real-world information consumption and sharing behaviors, although previous research has shown a correlation between self-reported willingness to share news in online surveys and actual sharing behavior on social media platforms51. Finally, while our study presents real news headlines that replicate a common social media design, the results may not generalize beyond our relatively small selection of political news headlines.