Robot-assisted language learning (RALL) is attracting growing attention among researchers and teachers thanks to its suggested positive impact on children’s motivation and concentration in second language (L2) learning. However, the implementation in real classroom situations has not yet been widely investigated. Addressing this gap, this study explores the learning of French pronunciation with the help of a social robot as a part of the normal L2 classroom routine without the presence of a human teacher. A social robot has a capacity to interact autonomously or semi-autonomously with humans, and it has a physical embodied presence, unlike chatbots. Fifty Finnish primary school children, aged 10 to 12 years, participated in the pedagogical experiment, presented as a challenge game called Défi Domi that was a three-level game where learners were challenged by a set of pronunciation tasks. They learned by imitating the robot without explicit pronunciation rules. The design enabled exploration of the robot’s role in learning pronunciation and its effects upon learners’ persistence. The results showed that learners’ pronunciation improved during the experiment and that pupils’ active participation became a natural part of the learning process. The participants were persistent, and they enjoyed learning with the robot throughout the challenge. This study argues that gamified RALL designs can support interactive and autonomous learning when implemented in classrooms.