OBJECTIVE: To understand reasons for suboptimal and delayed uptake of antiretroviral therapy (ART) by describing the patterns of HIV testing and counselling (HTC) and outcomes of ART eligibility assessments in primary clinic attendees. METHODS: All clinic attendances and episodes of HTC were recorded at two clinics in Blantyre. A cohort of newly diagnosed HIV-positive adults (>15 years) was recruited and exit interviews undertaken. Logistic regression models were constructed to investigate factors associated with referral to start ART. Qualitative interviews were conducted with providers and patients. RESULTS: There were 2398 episodes of HTC during 18,021 clinic attendances (13.3%) between January and April 2011. The proportion of clinic attendees undergoing HTC was lowest in non-pregnant women (6.3%) and men (8.5%), compared with pregnant women (47.2%). Men had more advanced HIV infection than women (79.7% WHO stage 3 or 4 vs. 56.4%). Problems with WHO staging and access to CD4 counts affected ART eligibility assessments; only 48% completed ART eligibility assessment, and 54% of those reporting WHO stage 3/4 illnesses were not referred to start ART promptly. On multivariate analysis, HIV-positive pregnant women were significantly less likely to be referred directly for ART initiation (adjusted OR: 0.29, 95% CI: 0.13-0.63). CONCLUSIONS: These data show that provider-initiated testing and counselling (PITC) has not yet been fully implemented at primary care clinics. Suboptimal ART eligibility assessments and referral (reflecting the difficulties of WHO staging in primary care) mean that simplified eligibility assessment tools are required to reduce unnecessary delay and attrition in the pre-ART period. Simplified initiation criteria for pregnant women, as being introduced in Malawi, should improve linkage to ART.