The integration of Artificial Intelligence (AI) chatbots in Nigerian banking has transformed customer service delivery, yet rigorous empirical assessments of their communicative effectiveness remain scarce. This study examined the usage patterns of customised AI chatbots of Sterling Bank (Naya) and United Bank for Africa (Leo) by undergraduates at Bowen University, Iwo, Osun State;evaluated the operational/communicative performance of these chatbots; identified common issues encountered using them; and assessed their influence on customers’ perceptions of service quality. Anchored on the Uses and Gratifications Theory and Technology Acceptance Model, the study employed a mixed-methods design combining a survey among purposively selected 344 undergraduates and in-depth interviews with customer service officials of the selected banks. Findings reveal that information retrieval was the dominant purpose of chatbot use (47.1%). Misunderstanding or misinterpretation of queries accounted for 36.87% of all reported issues. However, response accuracy (27.1%) and contextual understanding (31.2%) emerged as critical weaknesses. Overall, 41.3% of respondents were often satisfied, 42.2% sometimes satisfied, and 11.7% expressed rare or no satisfaction. The researchers conclude that customised AI chatbots in Nigerian banking have established a functional but communicatively limited presence, requiring urgent improvements in natural language processing accuracy, contextual intelligence, and linguistic adaptability to Nigeria’s multilingual environment. Nigerian banks should institute periodic independent audits of chatbot communicative accuracy, build continuous feedback mechanisms for real-time learning, and invest in NLP models trained on Nigerian language data, including Pidgin English and indigenous language-influenced varieties.