Imagine predicting the impact of a marketing campaign before it even launches. That's the promise of CXSimulator, a cutting-edge AI framework designed to forecast user behavior and revolutionize how we assess the effectiveness of web marketing strategies. Traditional methods like A/B testing are costly and time-consuming. CXSimulator offers a powerful alternative by simulating customer journeys through the use of Large Language Models (LLMs). How does it work? The system transforms user behaviors, like viewing a product or using a coupon, into semantic vectors using the power of LLMs. It then learns to predict the transitions between these events, creating a dynamic map of customer experience. This allows marketers to simulate how users might react to new campaigns or products *before* they go live, eliminating the need for expensive real-world testing. This simulated environment enables marketers to get an early preview of campaign success metrics, such as conversion rates, allowing for data-driven refinements. Tested against real-world data from the Google Merchandise Store, CXSimulator showed impressive accuracy in predicting customer behaviors. What's more, in a study with marketing professionals, the simulator’s predictions aligned closely with the intuitions of seasoned experts. This innovative approach promises to streamline the marketing process, enabling more targeted and effective campaigns. While still in its early stages, CXSimulator tackles the core challenge of predicting user behavior in the complex landscape of online marketing. Future developments could include incorporating more domain-specific knowledge to handle nuanced scenarios, and improving the interpretability of the results to provide even more actionable insights for marketers.