In South Africa, screening for diabetic retinopathy (DR) is non-existent at the primary healthcare (PHC) level because of the absence of a screening programme. This leads to preventable vision loss. To describe the clinical characteristics and outcomes of eye screenings and subsequent referrals. Laudium Community Health Centre (CHC), a PHC facility in Tshwane. We conducted a cross-sectional study from February 2022 to August 2022. Individuals with diabetes were screened for eye complications using visual acuity testing, intraocular pressure measurement, and fundoscopy with a non-mydriatic digital fundus camera. Fundus images were analysed by an optometrist and an artificial intelligence (AI) programme. Demographic and clinical data were collected. A total of 120 participants were included, with the majority (60.7%) from Laudium CHC. Most participants (64.2%) were on oral agents, and 66.7% were women. The mean haemoglobin A1c (HbA1c) was 8.3%, with a median diabetes duration of 8 years. Artificial intelligence detected more glaucoma cases (17.5% vs 9.2%) and DR (23.3% vs 15.8%) compared to the optometrist. In contrast, the optometrist identified more cases of macula pathology (29.2% vs 19.2%). Participants (n = 79) were referred to an ophthalmologist for diagnosis confirmation and management. The study revealed that while DR was not highly prevalent among PHC patients with diabetes, there was a significant referral rate for other ocular complications. Artificial intelligence can enhance early detection and improve efficiency. The findings underscore the need to integrate diabetes eye screening programmes into PHC services for people living with diabetes.