Dentinogenesis imperfecta (DI) is the main orodental manifestation of osteogenesis imperfecta (OI) caused by COL1A1 or COL1A2 heterozygous pathogenic variants. Its prevalence varies according to the studied population. Here, we report the molecular analysis of 81 patients with OI followed at reference centers in Brazil and France presenting COL1A1 or COL1A2 variants. Patients were submitted to clinical and radiographic dental examinations to diagnose the presence of DI. In addition, a systematic literature search and a descriptive statistical analysis were performed to investigate OI/DI phenotype–genotype correlation in a worldwide sample. In our cohort, 50 patients had COL1A1 pathogenic variants, and 31 patients had COL1A2 variants. A total of 25 novel variants were identified. Overall, data from a total of 906 individuals with OI were assessed. Results show that DI was more frequent in severe and moderate OI cases. DI prevalence was also more often associated with COL1A2 (67.6%) than with COL1A1 variants (45.4%) because COL1A2 variants mainly lead to qualitative defects that predispose to DI more than quantitative defects. For the first time, 4 DI hotspots were identified. In addition, we showed that 1) glycine substitution by branched and charged amino acids in the α2(I) chain and 2) substitutions occurring in major ligand binding regions—MLRB2 in α1(I) and MLBR 3 in α2(I)—could significantly predict DI ( P < 0.05). The accumulated variant data analysis in this study provides a further basis for increasing our comprehension to better predict the occurrence and severity of DI and appropriate OI patient management.