AI Platform Competition and Revenue Model Choice in Online Content Markets

Generative AI platforms built on large language models (LLMs) are rapidly becoming de facto gateways to information for many users. This transformation raises some important questions about the sustainability of traditional content platforms and the evolution of digital business models. While some publishers engage in high-stakes legal battles over content rights and others negotiate licensing agreements, there remains a noticeable absence of rigorous economic analysis to guide these decisions and inform policy responses. This project develops an economic model to analyze competition between traditional content platforms and AI-powered alternatives. We study a market where two heterogeneous content platforms (differing in quality) compete with an LLM platform that has been trained on their content and offers perfect personalization to users. Content platforms can choose between subscription-based and advertising-based revenue models, while users are heterogeneous along two dimensions: their horizontal preferences for content and their sensitivity to advertising. The LLM platform's distinctive features (its ability to provide personalized content and its intermediate quality derived from training on existing content) create some novel competitive dynamics that depart from standard platform competition. Our preliminary analysis yields several counterintuitive insights. First, we demonstrate that LLM entry can generate Pareto improvements, simultaneously increasing user welfare and incumbent platform revenues. This can occur when the threat of AI competition induces traditional platforms to switch from subscription to advertising models, creating a more efficient segmentation where some users access ad-supported content while others subscribe to the LLM service. Second, we show that LLM-induced elimination of low-quality platforms can benefit both users and surviving high-quality platforms by reducing competitive pressure and raising average market quality. Third, we characterize how platforms' strategic responses (including aggressive pricing and revenue model switching) depend on the distribution of user types, advertising sensitivities, and quality differentials. Planned extensions include incorporating content licensing agreements between AI and traditional platforms, endogenizing quality investments, and analyzing mixed revenue models for the LLM platform. This research has immediate policy relevance for ongoing regulatory debates about AI platform competition and content licensing requirements. Our framework can inform key decisions about whether AI platforms should compensate content creators for training data and whether AI represents a monopolistic threat or a welfare-enhancing competitive force. By providing an economic analysis of AI platform competition with endogenous revenue model choice, this project aims to guide both business strategy and regulatory design during this pivotal transformation of online content markets.