The aims of this study were to identify characteristics of the top-cited articles in problem-based learning (PBL) and assess the quality of evidence provided by these articles. The most frequently cited articles on PBL were searched in April 2015 in the Science Citation Index Expanded database (List A) and Google Scholar database (List B). Eligible articles identified were reviewed for key characteristics. The Oxford Centre for Evidence-Based Medicine guidelines were used in assessing the level of evidence. The number of citations varied (62 to 923 on List A and 218 to 2,859 on List B). Countries that contributed the majority of articles in both lists were the United States, Netherlands, United Kingdom, and Canada. No significant correlations were found between number of citations and number of years since published (p=0.451), number of authors (p=0.144), females in authorship (p=0.189), non-medical authors (p=0.869), number of institutions (p=0.452), and number of grants (p=0.143), but a strong correlation was found with number of countries involved (p=0.007). Application of the Oxford hierarchy of evidence showed that 36 articles were at levels 4 and 5 of evidence. This study found that research articles represented approximately one-third of PBL articles assessed and reported mainly on questionnaire-based studies. The most highly cited articles occupied top-ranking positions in the journals in which they were published. The lower level of evidence observed in most top-cited articles may reflect the significance of innovative ideas or content of these articles. These findings have implications for dental educators and dental researchers.