Can AI Improve Marketing Communications for Small Businesses? Field Evidence from Early Adopters 

Abstract

Generative AI writing has the potential to improve marketing communications, especially for small businesses. Using field data from over three hundred thousand private mental health care practices, I measure how adopting generative AI for writing practice descriptions affects patient demand, and how practice owners' experience moderates the benefits from adoption. To recover practices' adoption of generative AI, I train a custom ChatGPT detector to classify whether therapists used ChatGPT to re write a two hundred word description of their practice on a large patient search platform. I then identify the effect of adoption on patient demand in a difference-indifferences analysis. Demand increases on average after adopting ChatGPT, but decreases for inexperienced practice owners. Inexperienced owners sharply increase prices after adoption driving the reduction in demand, before reversing price increases later after demand falls. The data is consistent with a behavioural mechanism where inexperienced owners initially overestimate the quality of AI-generated content, leading to a short-run complementarity between generative AI and owner experience.

Keywords: Generative AI, Marketing Communications, Psychiatric Care, Natural Experiment

JEL Classification: O33, M31, I11, D83, M37