INTRODUCTION: Poor client-provider communication is a critical barrier to long-term retention in care among people living with HIV. However, standardized assessments of this key metric are limited in Africa. We used the Roter Interaction Analysis System (RIAS) to quantitatively characterize patterns of person-centred communication (PCC) behaviours in Zambia. METHODS: We enrolled pairs of people living with HIV making routine HIV follow-up visit and their providers at 24 Ministry of Health-facilities supported by the Centre for Infectious Disease Research in Zambia in Lusaka province between August 2019 and November 2021. Client-provider encounters were audio-recorded and coded using RIAS by trained research staff. We performed latent class analysis to identify interactions with distinctive patterns of provider PCC behaviours (i.e. rapport building, person-centred counselling, PCC micropractices [e.g. brief empathy statements], assessing barriers to care, shared decision-making and leveraging discretionary power) and compared their distribution across client, provider, interaction and facility characteristics. RESULTS: We enrolled 478 people living with HIV and 139 providers (14% nurses, 73.6% clinical officers, 12.3% were medical officers). We identified four distinct profiles: (1) "Medically Oriented Interaction, Minimal PCC Behaviours" (47.6% of interactions) was characterized by medical discussion, minimal psychosocial/non-medical talk and low use of PCC behaviours; (2) "Balanced Medical/Non-medical Interaction, Low PCC Behaviours" (21.0%) was characterized by medical and non-medical discussion but limited use of other PCC behaviours; (3) "Medically Oriented Interaction, Good PCC Behaviours" (23.9%) was characterized by medically oriented discussion, more information-giving and increased use of PCC behaviours; and (4) "Highly person-centred Interaction" (7.5%) was characterized by both balanced medical/non-medical focus and the highest use of PCC behaviours. Nurse interactions were more likely to be characterized by more PCC behaviours (i.e. Class 3 or 4) (44.8%), followed by medical officers (33.9%) and clinical officers (27.3%) (p = 0.031). Longer interactions were also more likely to integrate more PCC behaviours (p < 0.001). CONCLUSIONS: PCC behaviours are relatively uncommon in HIV care in Zambia, and often limited to brief rapport-building statements and PCC micropractices. Strengthening PCC, such as shared decision-making and leveraging discretionary power to better accommodate client needs and preferences, may be an important strategy for improving the quality in HIV treatment programmes.