In order to solve the problems of user fatigue caused by low relevance of push content and lagging information in online shopping, the work aims to optimize the personalized recommendation ability of the push system, and explore the feasibility of the interaction mode under the AIGC recommendation algorithm. Based on the rooted theory, the user association attributes were obtained and sorted, and then the users' demand information and product feature images were processed by the AIGC large model. Finally, the push model of "User-Product Point-To-Point Matching" based on AIGC was established. With the online shopping process of water cup as an example, the comparison experiment was carried out for verification. The experimental results showed that this model could improve user satisfaction in the aspects of information selection layer and deep search. The introduction of AIGC into the push service industry and the implementation of recommendation through the mode of "pre-training language model + fine-tuning" can further improve the personalized push service of "user-product", which is of great significance in e-commerce, social media, online education services and other fields.