With the rapid advancement and widespread adoption of artificial intelligence (AI), AI-powered personalized product recommendation (AI-PPR) has become a core tool for enhancing user experience and driving monetization on short-form video platforms, fundamentally reshaping consumer behavior. While prior research has largely focused on impulse buying intention (I-BI)—purchases triggered by emotional and sensory stimuli—there remains a lack of systematic exploration of spontaneous buying intention (S-BI), which emphasizes rational and cognitively driven decisions formed in unplanned contexts. Addressing this gap, this study integrates the Technology Acceptance Model (TAM) with a perceived evaluation perspective to propose and validate a dual-mediation framework: “AI-PPR → Perceived Usefulness/Perceived Trust → S-BI”. Using a large-scale survey of Generation Z users in mainland China (N = 754), data were analyzed via SPSS 26.0, including reliability and validity tests, regression analysis, and Bootstrap-based mediation analysis. The results indicate that AI-PPR not only has a significant positive direct effect on S-BI but also exerts strong indirect effects through perceived usefulness and perceived trust. Specifically, perceived usefulness accounts for 35.17% and perceived trust for 31.18% of the mediation, jointly constituting 66.35% of the total effect. The findings contribute theoretically by extending the boundary of purchase intention research, differentiating rational S-BI from emotion-driven impulse buying, and enriching the application of TAM in consumption contexts. Practically, the study highlights the importance for short-form video platforms and brand managers to enhance recommendation transparency, interpretability, and trust-building while pursuing algorithmic precision, thereby fostering rational spontaneous buying and achieving a balance between short-term conversions and long-term user value.