Auctions with LLM Summaries

The advent of large language model (LLM) technology has the potential to change the user experience of online services such as internet search, online recommendations (Geng et al., 2022), or shopping (Fan et al., 2023). For example, search platforms and apps, e.g., Microsoft Bing (Microsoft, 2023) and Google Search (Google, 2023), have already experimented with generative AI tools to provide augmented search summarization to facilitate users' search experience. Such summarization (e.g., based on retrieval augmented generation RAG (Lewis et al., 2020)) can sometimes provide an efficient way for users to gain useful information in a more condensed space. For queries of a commercial nature, search platforms respond with relevant online advertising. Online search advertising has provided a means not only to connect buyers and sellers, but also to support free internet services to users. Given the exciting potential of LLMs to summarize multiple sources of content and provide a succinct and informative output, it is natural to ask how LLM technology can help improve online advertising. In the ever-evolving landscape of online advertising, auction design has been a critical component towards improving the effectiveness and efficiency of ad delivery. A well-designed auction mechanism not only provides revenue for the platform but also ensures relevancy and value for users and advertisers alike.