Linking mitochondrial DNA (mtDNA) mutations to patient outcomes remains a formidable challenge. The multicopy nature and potential heteroplasmy of the mitochondrial genome, differential distribution of mutant mtDNAs among various tissues, genetic interactions among alleles, and environmental effects currently hamper clinical efforts to diagnose mitochondrial disease. Multiple sequence alignments are often deployed to estimate the potential significance of mitochondrial variants. However, factors including sample set bias, alignment errors, and sequencing errors currently limit the utility of multiple sequence alignments in pathogenicity prediction. Here, we describe an approach to assessment of site-specific conservation and variant acceptability that is reliant upon ancestral phylogenetic predictions and minimizes current alignment limitations. Using the output of our approach, we deploy machine learning in order to predict the pathogenicity of thousands of human mtDNA variants not yet linked to disease. Our work demonstrates that a substantial portion of mtDNA variants not yet characterized as harmful are, in fact, likely to be deleterious. Our findings will be of direct relevance to those at risk of mitochondria-associated illness, but the general applications of our methodology also extend beyond the context of mitochondrial disorders.