Using LLMs to compare 500 Pages of Macro Research (with citations) (2026macro.vercel.app)

A recent project utilized GPT-5.1 to analyze and compare 500 pages of macroeconomic research from major investment banks, resulting in a site that synthesizes various reports while providing traceable citations. The process involved a comprehensive six-pass pipeline: summarizing each report, defining comparison axes, extracting evidence with page references, and creating easily navigable HTML tables. Impressively, a spot-checking of 100 citations revealed a remarkably low error rate—indicating that the model can be used effectively for accurate research synthesis. This endeavor represents a significant advancement in the AI/ML community, demonstrating that large language models can be leveraged for tasks typically fraught with citation errors and inconsistencies. By enforcing a format for citations and employing thorough verification methods, the project sets a foundation for enhancing trust in AI-generated content. The implications extend beyond this specific analysis, suggesting that LLMs may play a crucial role in future research synthesis, literature reviews, and due diligence tasks by providing clear, source-cited claims, thus enabling deeper insights into complex datasets.