What does generative AI do to work?

This site collects empirical studies of generative AI in the workplace and puts their results on a common scale: speed effects as log ratios of task time or output, and quality effects as Hedges’ g. Each effect is linked to an O*NET occupation or work activity, so the evidence can be read against the structure of the labor market.

Browse the coded studies below, and click a title to see what was extracted from the paper and why. The imputation runs extend these observations across the O*NET graph; the method page explains how.