Using AI-generated mock jurors could be quicker, and certainly cheaper, but is it valid and reliable? A recent research analysis (Schweitzer & Salerno, 2025) says “maybe not.” Comparing the predictions made by commercially available AI-generated “mock jurors,” the researchers looked at how Large Language Models (LLMs) fared in predicting outcomes in two cases. They concluded: 

Not only do the predictions differ widely from human decisions, they differ among each other and even within themselves. These results should come as no surprise to those who are familiar with the workings of LLMs, which were never trained on nor intended to replace human judgment. Artificial intelligence – even when trained on mass quantities of actual human behavior – still struggles to predict what shows we might want to watch next, what songs we want to hear, who we want to date, and what products we will buy.