UCLA

When OpenAI launched ChatGPT, it did not immediately ship a mobile app. In the intervening months, dozens of third-party "wrapper" applications resold ChatGPT through OpenAI's public API, sharing its branding and competing for its App Store keywords—and by the official launch, imitators accounted for the vast majority of category downloads. The conventional view treats such imitation as a threat to the innovator. This dissertation argues the opposite: in a market where the imitated firm also sells the imitated technology as an API, imitators can function as contributors to the incumbent's long-run market power. I develop this thesis in three building blocks. Chapter 2 documents, with a daily panel of AI chatbot apps and keyword-level advertising data, that wrappers expand the total market, monetize underserved segments through flexible pricing, and coincide with a search-based entry barrier against independent rivals. Chapter 3 estimates a random-coefficients nested-logit demand system and finds that demand lost by a GPT-ecosystem product diverts to other GPT-ecosystem products and to the outside option, and essentially none of it—zero to four decimal places—to rivals running their own foundation models, which return only 0.13 percent of their own lost demand to the GPT ecosystem; removing the wrapper fringe would shrink the category by a third while raising the official app's profit one percent. Only a small share of that disadvantage is the wrappers' contemporaneous presence; most of it is habit accumulated during the fourteen months before any independent app established a presence in these markets—which moves the mechanism out of the static product space and into a stock, and is why the third chapter is followed by a dynamic one. Chapter 4 builds a dynamic model in which the platform's posted API price clears two markets at once—an enterprise token market where it competes with rival APIs, and the downstream wrapper channel that builds its consumer brand-habit stock—and shows that the observed 200x API price decline is competition rather than generosity, that the wrapper-to-brand spillover is point-identified in the awareness channel and set-identified above it, and that tolerating imitators was worth $21–86M to the platform while raising total welfare, though the consumer share of that gain is an order of magnitude smaller than the producer share. Together the chapters complicate the premise that open API access is inherently procompetitive: open access can entrench an incumbent by outsourcing brand proliferation to imitators the platform simultaneously supplies.