Background <p>Mitochondrial dysfunction contributes to poor embryo quality and recurrent assisted reproductive technology (ART) failure. Mitochondrial transplantation (MIT), which involves supplementing oocytes with exogenous mitochondria, has been proposed as a novel strategy to improve ART outcomes. However, both its clinical efficacy and safety remain unclear.</p> Methods <p>In this single-center trial, 151 patients with a history of ≥ 2 failed ART cycles provided 1178 metaphase II (MII) oocytes. Sibling oocytes were randomized 1:1 to receive autologous bone marrow mesenchymal stem cells (BMSCs) mitochondria co-injection during intracytoplasmic sperm injection (ICSI) or standard ICSI. The primary outcome was the rate of day-3 good-quality embryos.</p> Results <p>MIT significantly accelerated early embryonic cleavage at the 3-cell stage and 5-cell stage, but this morphokinetic alteration did not translate into improvements in good-quality embryo rate, clinical pregnancy rate, or live birth rate. Long-term follow-up of 23 live births revealed no adverse effects, with all offspring exhibiting normal growth and development. Exploratory analysis revealed that oocytes yielding ≥ 70% transferable embryos after MIT harbored an elevated higher burden of medium frequency (0.05–0.5) mtDNA point mutations.</p> Conclusions <p>While autologous BMSCs-MIT transiently alters early cleavage kinetics, it does not demonstrate a clinical advantage in unselected patients with recurrent ART failure. Nevertheless, its observed safety profile and the identification of mtDNA mutation burden as a potential predictive biomarker provide a foundation for shifting future MIT research from a universal approach toward precision application in molecularly stratified populations.</p> <p><i>Trial registration</i> ClinicalTrials.gov registration: NCT03639506</p>