GENERATIVE ARTIFICIAL INTELLIGENCE EQUALISES LOW-DIGITAL MICROENTERPRISE CAPABILITY

Abstract

This study investigates whether a generative AI copilot can raise marketplace performance for micro and small enterprises and whether the gains concentrate among lower-ability sellers. Across a 12-week randomized controlled field trial, 240 Manado sellers were evenly split between a high-traffic mall and a residential area and assigned either to business-as-usual or to an AI copilot that generated optimized titles, persuasive descriptions, relevant tags, and photo guidance for their TikTok videos. We tracked the click-through rate, conversion rate, average order value, buyer rating, and an internal listing-quality index, and estimated causal effects using difference-in-differences models with seller-clustered standard errors. Compared with the control, the AI treatment increased click-through and conversion rates, raised the average order value by approximately IDR 19,000, improved ratings, and enhanced listing quality, with more potent effects in the residential segment. Mediation analysis showed that AI primarily improved listing quality; the resulting quality gains translated into higher conversion rates, and the indirect effect was statistically significant. The approach is scalable as a low-cost, platform-embedded marketing aid in urban markets.