Abstract Study question How does the accuracy of an artificial intelligence (AI)-enhanced semen analysis system compare to conventional semi-automated methods when assessing semen quality? Summary answer Compared to conventional semi-automated methods for sperm quality assessment, the AI-enhanced system tends to overestimate progressive sperm motility and normal sperm morphology. What is known already Male factor infertility, specifically poor sperm quality, accounts for over 40% of all infertility cases. Accurate semen analysis is thus a critical step in guiding appropriate treatment strategies in male reproductive health. Past decades have witnessed the development of semi-automated systems for semen analysis, overcoming the limitations of traditional microscopy. While these systems offer precise data on kinetics and concentration, manual intervention remains necessary for assessing certain parameters, potentially affecting reproducibility. AI-based systems have been recently introduced into clinical practice, promising more objective and standardized analysis. Yet, the extent to which these methods surpass conventional approaches remains to be established. Study design, size, duration This is a single-centre, prospective, consecutive, paired study, conducted from February to September 2022. The study included 84 sperm samples from patients undergoing medically assisted reproduction. Participants/materials, setting, methods All sperm samples were analyzed using the AI-enhanced automated system (LENSHooke X1 Pro, Bonraybio®, Fertil Ibérica) and conventional methods: CASA semi-automatized device (ISAS, Proiser), Diff-Quik staining for manual morphology assessment, and pH test strips. We measured and compared several semen parameters, including morphology (% normal sperm), concentration (million/mL), motility (% progressive, % non-progressive, % total motility), and pH, using both approaches. Spearman and Kruskal-Wallis tests were used for statistical analyses. Main results and the role of chance We observed significant differences among several semen parameters when comparing the AI-enhanced system and conventional approaches. Mean normal sperm morphology, as assessed by the AI-enhanced automated device, was significantly higher than that evaluated by conventional microscopy (4.62% versus 3.07%, p < 0.00001). We also observed significant differences for motility and pH parameters (57.98 versus 25.66, p = 0.0001 and 7.86 versus 8.38, p = 0.0001, respectively). Correlations between variables assessed using both methods also varied. Sperm concentration showed a strong correlation between the AI-enhanced and conventional systems (Spearmann’s Rho=0.828), however a weak correlation was observed for total and progressive motility, as well as the percent of immotile sperm (69.54% versus 28.98% R2=0.541; 57.98% versus 25.66% R2=0.566, and 30.45% versus 47.12%, R2=0.541, respectively). There was no significant correlation for non-progressive motility between the two methods (6.53% versus 11.60%, R2=0.157). Limitations, reasons for caution This study exclusively compared two analysis methods among numerous AI-enhanced systems available on the market. Given the rapidly evolving landscape of AI technologies, caution is warranted when generalizing these findings to other AI models, as their performance may vary. Wider implications of the findings Observed differences between the AI-enhanced system and conventional methods require careful consideration. While fully automated semen analysis systems are designed to reduce subjectivity, extensive comparative research is crucial. A comprehensive market assessment will be essential for safeguarding patient treatment outcomes and advancing AI capabilities in semen analysis. Trial registration number not applicable