Agentic Recommendation Markets: LLM Agents Shift Power from Platforms to Users
A new arXiv preprint (2607.25253) introduces the concept of agentic recommendation markets, where LLM-based user agents allow users to specify needs before choosing a platform, forcing platforms to compete for attention. In controlled experiments across three product domains, researchers found that this user-centric model expands the pool of relevant items but creates a tension between access and attention. Platforms respond strategically, with selectively positive explanations occupying 73–78% of first-ranked positions. The study highlights how agentic markets reshape the dynamics of online recommendation, shifting from platform-centric curation to user-driven discovery.
Key facts
- arXiv preprint 2607.25253 introduces agentic recommendation markets
- LLM-based user agents enable users to specify needs before choosing a platform
- Platforms compete for user attention in this new setting
- Experiments conducted across three product domains
- User-centric recommendation expands opportunity for relevant items
- Broader participation does not translate directly into effective exposure
- Selectively positive explanations occupy 73–78% of first-ranked positions
- Competition triggers strategic play by platforms
Entities
Institutions
- arXiv