When Personalization Expands Markets: Evidence from a Field Experiment on Serialized Content

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Prior research on recommendation systems (RSs) largely assumes that personalization redistributes fixed aggregate demand across items rather than expanding total consumption. We challenge this assumption by theorizing consumption architecture as a boundary condition determining whether recommendations generate redistributive or expansionary effects. Using a large-scale randomized field experiment on a serialized digital content platform, we compare popularity-based and personalized recommendation interfaces among 400,000 users and 714 titles over 91 days. Personalized recommendations significantly increase total engagement, paid consumption, and adoption of new titles, while market concentration simultaneously declines. This joint pattern cannot arise under strict zero-sum substitution and instead indicates market expansion. We argue that serialized content platforms attenuate substitution constraints through low per-episode commitment costs, parallel consumption capacity, and temporal slack between release cycles, enabling personalization to activate latent demand. Our findings contribute a consumption architecture perspective to ongoing debates about the distributional consequences of RSs.