AI Exposure and Housing Markets

Abstract

Generative AI arrived just as the pandemic had pushed housing demand away from expensive, supply constrained metros. We ask whether the period after Chat-GPT extended that dispersion or reversed it. Using a quarterly panel of 211 U.S. metros from 2017Q1 to 2025Q4, we relate Zillow and FHFA house price growth to AI Geography Exposure weighted by 2019 population, with the 2020 remote work break timed separately. AI exposed metros grow faster after 2023: a one standard deviation increase in AIGE is associated with 3.6% higher cumulative ZHVI growth, about $9,400 at the sample median 2023 home value. The premium also appears in FHFA prices and rents, concentrates in supply inelastic metros, and is reproduced by a predetermined LLM task suitability index. A pandemic WFH placebo has the opposite sign. IRS migration data show reduced outmigration from exposed metros, and Revelio LinkedIn data show rising AI hiring there. The early generative AI period appears to have reinforced the residential value of dense, supply constrained high skill labor markets, reversing the spatial direction of the pandemic.

Keywords: AI exposure, housing prices, metropolitan areas, remote work, urban rebound

JEL Classification: R31, J23, O33, R11