The present study aims to develop a model of the impact of artificial intelligence recommender systems on online impulse buying, with the mediating role of emotional arousal and the moderating role of self-regulation. In terms of purpose, this research is applied-developmental; in terms of implementation, it is a mixed-methods study (qualitative–quantitative); and in terms of nature, it follows a sequential exploratory design.In the qualitative phase, the statistical population consisted of 18 online shoppers, who were selected through purposeful sampling using a snowball strategy. In the quantitative phase, the statistical population included 378 internet users residing in Tehran, who were selected through purposive convenience sampling.Data collection was carried out through semi-structured interviews in the qualitative phase and through a questionnaire in the quantitative phase. For analyzing the qualitative data, the grounded theory method and MAXQDA 2024 software were used, while SPSS and SmartPLS 4 were employed for the quantitative analysis.The qualitative findings revealed 25 concepts and 6 main categories. Among these, perceived urgency and emotional personalization showed the highest frequency in participants’ impulse buying experiences. The quantitative results indicated that perceived urgency has a positive and significant effect on AI recommender systems. Furthermore, the three variables of perceived urgency, emotional personalization, and surprising novelty had a positive and significant effect on arousal. Arousal, in turn, had a positive and significant effect on impulse buying.The findings also showed that arousal plays an important mediating role in the relationship between perceived urgency, emotional personalization, and surprising novelty and impulse buying. Overall, recommender systems primarily generate positive arousal through creating a sense of urgency and emotional personalization, which in turn leads to impulse buying, while high self-regulation plays a protective role