Introduction The objective of this study was to assess an original learning intervention to train students and paediatric dentistry teachers in radiographic diagnostic accuracy of pulpo‐periodontal complications in primary molars. Materials and Methods The learning intervention was based on 250 different randomly ordered radiographs of primary molars within three quizzes (A, B and C) for 5 sessions (S): quiz A (50 X‐rays), B and C (100 X‐rays) were, respectively, completed in S1 to assess the extent of agreement with 5 experts' diagnoses, in S2 and S3 (B at days 8 and 23) and in S4 and S5 (C at days 90 and 105). During S1 and at the end of S3 and S5, the participants (48 students and 16 teachers) were informed of correct diagnoses. A satisfaction questionnaire was completed by all the students. Alongside the descriptive analyses, generalised linear mixed model (GLMM) analyses assessed the odds of participants' correct diagnosis over the study duration. Results At S1, the odds of diagnostic accuracy among students were significantly lower than those among the teachers. After receiving feedback at S1, GLMM analyses showed that among all the participants, accuracy improved over time with the odds of correct diagnoses higher in S2‐5 than in S1; and there were similar increases across sessions between teachers and students, except in S3, where the improvement among teachers tended to be greater than that among the students. All students were satisfied though one‐third reported that quizzes with 100 radiographs felt too long. Conclusion The online case‐based learning was a good training format for dental education.