The CRISPR-Cas9 system represents a transformative therapeutic platform for monogenic blood disorders such as sickle cell disease (SCD) and β-thalassemia. However, clinical translation is limited by safety concerns related to unintended off-target genomic alterations. This systematic review and meta-analysis quantified the prevalence and evaluated the severity of CRISPR-Cas9 off-target effects in hematopoietic models of blood-related genetic diseases. Seventy-eight studies were included in qualitative synthesis, and five comparative studies were analysed using a random-effects meta-analysis. High-fidelity Cas9 variants significantly reduced off-target risk compared with wild-type SpCas9 (pooled OR = 0.15, 95% CI: 0.08–0.29), corresponding to an 85% reduction in off-target odds. Substantial heterogeneity was observed (I² = 76.3%), primarily attributable to differences in off-target detection methods, as variations in methodological sensitivity can systematically influence the number and frequency of detected off-target events, contributing substantially to the observed heterogeneity. Subgroup analyses showed greater specificity improvements when assessed using unbiased genome-wide methods (OR = 0.08) compared with targeted sequencing (OR = 0.23). Although high-fidelity variants markedly improve safety profiles, methodological variability and limited long-term in vivo data remain important considerations. These findings provide quantitative evidence supporting the potential use of high-fidelity nucleases and comprehensive genome-wide assessment to enhance the safety of CRISPR-based therapies for hemoglobinopathies.