This project compares stock recommendations generated by large language models against those issued by human equity analysts, testing whether accuracy and effectiveness depend on what kind of input each is given — structured financial statements, unstructured news and filings text, or a combination of both.

The goal is to identify the conditions under which generative AI recommendations match or diverge from human analyst judgment, and what that implies about where LLMs are (and aren’t yet) reliable substitutes for professional financial analysis.