Abstract Study question Can the use of artificial intelligence (AI) enhance the prediction of mature oocytes (MII) from Cumulus-Oocytes-Complexes (COC) significantly surpassing the accuracy of well-trained embryologists? Summary answer The development of a COC classifier, an AI-based tool, improves the prediction of mature oocytes from COCs graded as 3 and 4 before oocyte denudation. What is known already Determining the maturity of Metaphase II oocytes before denudation is a challenge in assisted reproduction due to the presence of a dense cumulus mass surrounding the oocyte and obstructing proper manual scoring, a critical step in intracytoplasmic sperm injection (ICSI) due to risk of granulosa cells removal from immature oocytes. Although there is a certain correlation between COC morphology and nuclear oocyte maturity, it is not 100% accurate. It is reported that approximately 90% of MII oocytes are expected in expanded COCs (grade 1), while in COCs grade 4, more immature oocytes are predicted and only 25% are MII oocytes. Study design, size, duration 931 COC's from 145 patients were submitted to COC grading by both, manual and a new “COC classifier” from October to January 2023. 4 groups were formed according to Ebner criteria, 2008. Grade 1: 192 COC's (suspected mature), fluffy and radiant corona and cumulus; Grade 2: 318 COC's showing fluffy cumulus (suspected mature); Grade 3: 208 COC's with radiant corona (expected mature) and grade 4: 213 COC's (suspected immature) with dense cumulus without visible oocyte. Participants/materials, setting, methods Prospective study included 145 patients (39.3 years ±3.2) from September to December 2023. After COC obtention, excess of granulosa cells, dark and blood cells were removed. 3 hours after egg collection, COCs were photographed and scored manually by 9 well trained embryologists. Before denudation, obtained COCs were classified in 4 morphological stages according to Ebner et al, 2008. Finally, the taken photos from COC's were analysed by the COC classifier, a sypervised deep learning algorithm. Main results and the role of chance After COC's classification as following: Group 1 (suspected mature), fluffy and radiant corona and cumulus with visible oocyte; Group 2, COC's showing dense corona (oocyte clearly visible) but fluffy cumulus (suspected mature); Group 3, COC's with radiant corona (oocyte visible, expected mature); Group 4, COC's (suspected immature) with dense corona and cumulus without visible oocyte. The new software presented an MII oocyte accuracy of 79.16% for group 1, 87.1% for group 2, 83.17% for group 3, and 73.23% for group 4. In comparison, manual COCs grading by nine experienced embryologists yielded accuracy rates of 73.5% MII oocytes in group 1, 79.5% for group 2, 50.3% for group 3, and 46.9% for group 4. Statistical analysis, using the chi-square test, was performed to compare the accuracy rates between the COC classifier and manual grading. The results revealed significant differences in favor of the COC classifier when compared to manual method for the groups 3 and 4 (p < 0.001). These results indicate a superior performance of the COC classifier in predicting MII oocytes for groups with highest rates of inmture oocytes. However, for groups 1 and 2, there were no statistically significant differences between the COC classifier and manual grading. Limitations, reasons for caution Further validation through a large-scale, prospective, randomized study is necessary to confirm the efficacy and reliability of the new COC classifier, particularly in patients with oocyte cohorts enriched with grade 3 and 4 COCs. Ongoing research is currently being conducted in our laboratory. Wider implications of the findings The new COC classifier, an AI-based tool, developed by our group, offers a non-invasive and efficient method in predicting mature oocytes before denudation by identifying more MII oocytes from patients’ cohorts enriched with grade 3 and 4 COC's, postponing denudation and injection timing or destinating them to conventional IVF Trial registration number 5678765