Aim To evaluate improvement in knowledge and clinical behaviour among healthcare professionals after attendance at paediatric epilepsy training (PET) courses. Method Since 2005, 1‐day PET courses have taught evidence‐based paediatric epilepsy management to doctors and nurses in low‐, middle‐, and high‐income countries. A cohort study was performed of 7528 participants attending 252 1‐day PET courses between 2005 and 2020 in 17 low‐, middle‐, and high‐income countries, and which gathered data from participants immediately after the course and then 6 months later. Training outcomes were measured prospectively in three domains (reaction, learning, and behaviour) using a mixed‐methods approach involving a feedback questionnaire, a knowledge quiz before and after the course, and a 6‐month survey. Results Ninety‐eight per cent (7217 of 7395) of participants rated the course as excellent or good. Participants demonstrated knowledge gain, answering a significantly higher proportion of questions correctly after the course compared to before the course (88% 47 883 of 54 196, correct answers/all quiz answers, vs 75% 40 424 of 54 196; p < 0.001). Most survey responders reported that the course had improved their epilepsy diagnosis and management (73% 311 of 425), clinical service (68% 290 of 427), and local epilepsy training (68% 290 of 427). Interpretation This was the largest evaluation of a global epilepsy training course. Participants reported high course satisfaction, showed knowledge gain, and described improvements in clinical behaviour 6 months later. PET supports the global reduction in the epilepsy ‘treatment gap’ as promoted by the World Health Organization. Short course training, particularly in low‐income settings, is rarely carefully evaluated. This study reports the findings, using a well‐established evaluation framework, of 250 1‐day paediatric epilepsy training courses delivered in 17 countries over 15 years. Participants conveyed high course satisfaction, demonstrated a significant improvement in knowledge after the course, and reported positive changes in clinical practice 6 months later. This original article is commented by Wiebe on pages 967–968 of this issue.