Against the backdrop of rising demand for high-quality academic writing among normal university students, cognitive inertia (e.g., procrastination, superficial thinking) in graduation thesis writing has become a key challenge. This study examined the application of an AI-driven intelligent feedback system (integrating natural language processing and learning analytics) in thesis writing of 100 senior students from Yunnan Normal University, and its intervention on cognitive inertia. The system provides real-time, multi-dimensional feedback (grammar correction, content optimization, logical guidance).A 12-week controlled experiment divided participants into an experimental group (n=50, using the AI system) and a control group (n=50, receiving only traditional teacher feedback). Data were collected via thesis quality scores, system logs, and semi-structured interviews.Results showed the experimental group’s thesis scores were 15.2% higher than the control group (p<0.01), with significant improvements in literature citation depth and argumentation logic. The system intervenes cognitive inertia through: 1) direct mechanism (shortening cognitive correction cycle, reducing average modification response time from 48h to 6h, activating metacognitive monitoring); 2) indirect mechanism (enhancing self-efficacy by 23.7%, mitigating writing anxiety via task decomposition).Findings indicate the AI system effectively improves writing quality and intervenes cognitive inertia, offering a new approach for normal education writing teaching.