Introduction: Mitochondrial diseases caused by mtDNA have no effective cures. Recently developed DddA-derived cytosine base editors (DdCBEs) have potential therapeutic implications in rescuing the mtDNA mutations. However, the performance of DdCBEs relies on designing different targets or improving combinations of split-DddA halves and orientations, lacking knowledge of predicting the results before its application. Methods: A series of DdCBE pairs for wide ranges of aC or tC targets was constructed, and transfected into Neuro-2a cells. The mutation rate of targets was compared to figure out the potential editing rules. Results: It is found that DdCBEs mediated mtDNA editing is predictable: 1) aC targets have a concentrated editing window for mtDNA editing in comparison with tC targets, which at 5'C8-11 (G1333) and 5'C10-13 (G1397) for aC target, while 5'C4-13 (G1333) and 5'C5-14 (G1397) for tC target with 16bp spacer. 2) G1333 mediated C>T conversion at aC targets in DddA-half-specific manner, while G1333 and G1397 mediated C>T conversion are DddA-half-prefer separately for tC and aC targets. 3) The nucleotide adjacent to the 3' end of aC motif affects mtDNA editing. Finally, by the guidance of these rules, a cell model harboring a pathogenic mtDNA mutation was constructed with high efficiency and no bystander effects. Discussion: In summary, this discovery helps us conceive the optimal strategy for accurate mtDNA editing, avoiding time- and effort-consuming optimized screening jobs.