Background Posture is assessed clinically and used to guide treatment of low back pain. Collectively, the relevance of posture and clinical postural assessments have come under scrutiny. This study aimed to determine (a) the intra-rater and inter-rater reliability of visual assessments of lumbar lordosis, and (b) the agreement between visual and direct postural assessments. Methods Ten physiotherapists visually assessed the lumbar lordosis from 3D scans of 50 asymptomatic participants, and 15 duplicates, using a grading scale of deviations (range: 0 = normal to 3 = severe). Lumbar lordosis angle was directly assessed using the Vitus Smart 3D whole body scanner. Cohen's Kappa was used to determine the intra-rater and inter-rater reliability of visual assessments, with polyserial correlation (ps) used to determine the agreement between visual and direct assessments. Results Overall, 93% and 83% of all intra-rater and inter-rater differences in visual assessments were within a single grade point, respectively. The intra-rater and inter-rater reliability of visual assessments was moderate (κ (95%CI): 0.56 (0.45, 0.67)) and slight (κ (95%CI): 0.13 (0.08, 0.19)), respectively. The agreement between visual and direct assessments was moderate (ps = −0.41, p = 0.04). Conclusion Visual assessments of lumbar posture demonstrated moderate repeatability and agreement with quantitative assessments. While agreement between assessors was slight, 83% of the visual ratings were within a single grade point, suggesting greater coherence among clinicians than our statistics suggested. As with any clinical assessments involving uncertainty, postural assessment should not solely guide treatment.