With the advent of rapid advancements in information technology, online learning has increasingly gained traction, prompting a multitude of colleges and universities to adopt innovative reforms in their educational delivery methods. This research undertakes the development of a college English online interactive learning system empowered by artificial intelligence. It primarily investigates the interactive behaviors of college students in online English learning environments. Initially, the study employs the K-means algorithm to categorize students based on their distinct learning behaviors. Subsequently, it leverages a recommendation algorithm to fine-tune the provision of English teaching resources online. Furthermore, the study integrates an enhanced Apriori algorithm to extract associations between student behaviors and learning outcomes. This analytical framework underpins the empirical evaluation of the system’s efficacy in actual educational settings, aiming to deepen the understanding of students’ online interactive behaviors. The findings reveal that over 95% of students consistently log into the learning platform on scheduled class days. The metrics for course recommendation success, such as the hit rate, the average inverse rank, and the range of normalized discounted cumulative gain, are reported between 18.37-21.94, 1.87-2.08, and 4.44-4.87, respectively. Additionally, system memory utilization remains stable at 57%-66%, corroborating the system’s operational effectiveness. This research contributes valuable insights and benchmarks for the development of robust online interactive learning systems.