In response to the lack of subjectivity, experience, and limitations in current marketing strategy research, this article applied a natural language processing (NLP) technology based on artificial intelligence (AI) to improve matching algorithms and enhance the personalization level of marketing strategies. Using natural language processing technology, text analysis and emotion recognition were performed on user demands. Based on the results of user demand analysis, a text semantic representation learning method based on the BERT (Bidirectional Encoder Representations from Transformers) model was proposed; deep reinforcement learning matching algorithms were used, and user behavior was combined with real-time feedback to optimize recommendation results. Personalized matching strategies were evaluated through indicators such as click through rate, conversion rate, and user satisfaction. The results showed that the satisfaction score obtained by using NLP method in the 18–25 age group was 6% higher than that of logistic regression algorithm.