A new review argues that artificial intelligence (AI) has transformed measurement in economics by making it possible to convert unstructured data into structured variables at low cost. This shift changes the focus from finding a scalable measure of a phenomenon to choosing among many plausible ones, which can lead to different empirical conclusions. The review provides guidance on how researchers should navigate this change and ensure credible inference with AI-generated variables. It emphasizes the importance of designing validation samples that anchor measurement to explicit criteria rather than relying on informal claims. The authors also discuss what happens when a random validation sample is unavailable. --- Why it matters: This matters because it highlights the challenges and opportunities in using AI for measurement, particularly in economics. Researchers need to be aware of these issues to ensure their findings are reliable and generalizable. Source: https://arxiv.org/abs/2608.23524