Human culture is the result of a unique cumulative evolutionary process. Despite the importance of culture for our species the social transmission mechanisms underlying this process are still not fully understood. In particular, the role of language – another unique human behaviour – in social transmission is under-explored. In this first direct, systematic comparison of demonstration vs language-based social learning, we ran transmission chains of participants (6- to 8-year-old children and adults) who attempted to extract a reward from a puzzle box after either watching a model demonstrate an action sequence or after listening to verbal instructions describing the action sequence. The initial seeded sequences included causally relevant and irrelevant actions allowing us to measure transmission fidelity and the accumulation of beneficial modifications through the lens of a subtractive ratchet effect. Overall, we found that, compared to demonstration, verbal instruction specifically enhanced the faithful transmission of causally irrelevant actions (overimitation) in children, but not in adults. Cumulative cultural evolution requires the faithful transmission of sophisticated, complex behaviour whose function may not be obvious. This indicates that, by supporting the retention of actions that appear to lack a causal function specifically by children, language may play a supportive role in cumulative cultural evolution.