With the development of digital advertising intelligence, programmatic advertising has become an important trend in advertising delivery, especially in the field of pharmaceutical advertising with strict supervision and outstanding commercial value, optimizing delivery effect and ensuring compliance are equally important. This paper focuses on the application of Multi-Armed Bandit (MAB) algorithm in programmatic pharmaceutical advertising. By constructing a joint profit function of CTR (click-through rate), CVR (conversion rate) and compliance penalty, the actual effects of three classic strategies, ε-greedy, UCB and Thompson Sampling, are compared. In the empirical analysis based on the US pharmaceutical advertising market data, the Thompson Sampling strategy performs best, with an average increase of more than 10% in CTR and CVR, and can effectively avoid compliance risks, this paper verifies the MAB algorithm's ability to balance the advantages between maximizing advertising revenue and compliance control through experimental design, algorithm modeling and multi-dimensional visualization. The research results show that the MAB strategy provides a data-driven, dynamically adaptive optimization path for programmatic pharmaceutical advertising, which has practical promotion and application value.