SMU SOE Seminar Series (July 15, 2026): Searching for Trade: Evidence from AI-Firm Interactions

TOPIC:

ABSTRACT

Search costs are a first-order barrier to trade. Online platforms promised to lower them by exposing firms to partners worldwide; yet, as exposure becomes abundant, the constraint shifts from being seen to being relevant. We examine whether generative AI addresses this constraint by studying an AI matchmaker on an international trade platform that reads each firm's search request and returns a tailored shortlist of relevant matches. The setting provides two rarely observed objects: the full text of every search and the algorithm's complete recommendations, including both the presented top three and the unpresented near-top matches. Linking these firm-AI interactions to customs microdata for Colombia and Uruguay, we investigate firms' search behavior and the effect of algorithmic matching. We find that AI recommendation significantly raises a firm's profile visits and trade value. The gain operates entirely through new partners rather than larger orders with existing ones, and concentrates in personalized, informative matches. The tool also stimulates additional searches, directing visits toward non-recommended firms. The findings provide evidence that AI matching can be valuable in trade search and its value is shaped by the relevance of matches rather than exposure alone.