Applying Large Language Models to Sponsored Search Advertising.

This article presents a human-in-the-loop large language model (LLM) framework designed to enhance search engine advertising (SEA) performance by generating ad content tailored to SEA contexts. The framework integrates keyword data from top organic search results and landing pages (LPs) to guide content generation using a plug and play language model (PPLM) approach, and includes a content scoring system alongside a predictive model for cost-per-click (CPC). Empirical field experiments with a midsized IT and SaaS provider and a business school demonstrate that this tailored LLM workflow outperforms human writers and state-of-the-art LLMs like GPT-4 and Google Gemini in impressions, clicks, and conversions, though higher content optimization may increase CPC. The research also identifies boundary conditions, such as potential performance declines when both ad and LP content are AI-optimized and the influence of budget constraints, highlighting the need to balance content quality and cost in SEA campaigns.