We consider a platform facilitating trade between sellers and buyers with the objective of maximizing consumer surplus. In many such platforms prices are set by revenue-maximizing sellers, but the platform may influence prices through its promotion policy (e.g., increasing demand to a certain product by assigning to it a prominent position on the webpage), and the information it reveals about the additional demand associated with being promoted. Identifying effective joint information design and promotion policies for the platform is a challenging dynamic problem as sellers can sequentially learn the promotion "value" from sales observations and update prices accordingly. We introduce the notion of confounding promotion polices, which are designed to prevent a Bayesian seller from learning the promotion value (at the cost of diverting consumers away from the best product offering). Leveraging this notion, we characterize the maximum longrun average consumer surplus that is achievable by the platform when the seller is myopic. We then establish that long-run average optimality can be maintained by optimizing over a class of joint information design and promotion policies under which the platform provides the seller with a (random) information signal at the beginning of the horizon, and then uses the best confounding promotion policy, which prevents the seller from further learning. Additionally, we show that myopic pricing is a best response to such a platform strategy, thereby establishing an approximate Bayesian Nash equilibrium between the platform and the seller. Our analysis allows one to identify practical long-run average optimal platform policies in a broad range of demand models and evaluate the impact of the search environment and the design of promotions on consumer surplus.