This paper examines the ability of a popular large language model—ChatGPT—to emulate the skills of equity analysts. We tackle this question using text transcripts from earnings conference calls. Our approach is two-fold. First we use the question-and-answer section of analyst to train ChatGPT regarding how analysts think about information relevant to their analysis. Next we apply what ChatGPT has learned to evaluate transcripts from 200 of the largest firms in the US by developing an Analyst Insight Score (AIS) for each firm-transcript pair. We find that our “machine- learned” AIS is consistent with observed behaviour from analysts including adjustments to price targets after earnings announcements. Moreover our AIS outperforms more traditional analyst metrics such as the Standard Unexpected Earnings (SUE). Finally we find that our AIS allows for the construction of portfolios that earn abnormal returns with respect to several standard asset pricing models.