With the widespread adoption of the internet and social media, the elderly has increasingly engaged in social interactions through these platforms. However, due to their limited ability to adopt digital technologies, they often face barriers posed by the digital divide while benefiting from digital advancements. As artificial intelligence (AI) continues to develop rapidly, social media platforms have become deeply integrated with AI technologies, significantly transforming user experiences and interaction patterns. For the elderly, AI-integrated social media platforms can expand access to information and enhance social interaction engagement. On the other hand, new forms of the digital divide may introduce additional challenges and concerns. This study examines the behavioural mechanisms underlying the elderly’s use of AI-integrated social media platforms and their participation in social interactions. It draws on the concept of the digital divide, the norm activation model, and social interaction theory. Through in-depth interviews with 12 elderly participants, the study applies coding and inductive analysis to develop an exploratory model within the stimulus-organism-response theoretical framework. The findings demonstrate that the digital divide significantly influences the elderly’s adoption of digital technologies and engagement in social interactions. This study not only enriches theoretical discussions in this field but also provides practical recommendations for social media practitioners and policymakers.