Human Detection of AI-Generated Consumer Reviews: An Eye-tracking Study

Sohyun Park (Ph.D. Student), Bartosz Wojdynski, Moses U. Okocha (Ph.D. Student), and Jiwon Kim (Ph.D. Student). "Human Detection of AI-Generated Consumer Reviews: An Eye-tracking Study,” paper presented at the 2025 AEJMC conference Communication Technology Division, August, San Francisco.

Abstract: Online information searching in relation to products and services increasingly requires humans to make decisions about whether user-generated content is authentic or synthetic. Considering this new challenge, an experimental study was conducted to examine how distinct message factors influenced user assessments of the authenticity of online consumer reviews on the Yelp platform. This study adopted a 2 (review text provenance: human-generated vs. AI-generated) × 2 (review valence: positive vs. negative) × 2 (image provenance: human-generated vs. AI-generated) × 2 (image presence: profile only vs. profile and product images) mixed factorial design, integrating eye-tracking technology. All variables were varied within subjects according to a fractional factorial design, and each participant was presented with a series of eight mocked-up Yelp reviews.