Who Uses AI? Platforms, Workforce, and AI Exposure

TL;DR

This paper critically examines how AI platform conversation logs are used to measure occupation exposure, revealing that these measures are influenced by platform user base rather than actual workforce exposure, and discusses the implications of measurement error.

Contribution

It formalizes the non-classical measurement error in AI exposure metrics, derives bounds for employment elasticities, and highlights biases in current measurement approaches.

Findings

Platform-based AI exposure scores are partly driven by user base rather than workforce.

Reweighting to BLS workforce shares significantly reduces estimates.

Bias in measurement understates substitution effects more than augmentation.

Abstract

A growing literature uses artificial intelligence platform conversation logs to measure occupation exposure. We show that these scores partly measure platform user base rather than the workforce. Holding outcome, sample, controls, and estimator fixed while varying only the platform input changes the post-ChatGPT employment coefficient by a factor of 1.9, and within-vendor consumer-versus-enterprise channels produce estimates that disagree in sign. Reweighting to Bureau of Labor Statistics workforce shares attenuates estimates by 42 to 93 percent. We formalize the non-classical measurement error, derive probability limits and partial-identification bounds for employment elasticities. The bias understates substitution more than augmentation.