Too Much of a Good Thing? Artificial Intelligence Usage and Analyst Forecast Accuracy

In this study, we examine whether AI helps or harms analyst forecasts. We find that there is an inverted U-shaped relationship between AI usage and analyst forecast accuracy. Specifically, moderate AI usage enhances analysts’ forecast accuracy, whereas excessive reliance attenuates it. Further mechanism analyses reveal that AI influences analysts’ forecasting behavior through both an information channel and a capability channel: 1) In environments with limited information, AI plays a role in expanding the information boundary, thus its impact on analysts’ forecast accuracy is more pronounced. Whereas in environments with abundant information, the marginal benefit of AI usage for forecast accuracy diminishes. 2) AI significantly improves the forecast accuracy of low-skilled analysts but exerts no statistically significant effect on their high-skilled counterparts. Finally, our findings remain robust to multiple robustness tests. Our study provides a human-machine integration perspective on the study of information dissemination in financial markets and analysts' predictive abilities, while also revealing that over-reliance on AI may lead to prediction errors. JEL classification : G14 G17 M41 O33