Educational escape rooms are taxing in terms of the time needed to design, create, conduct, and evaluate. Therefore, a high "return on investment" is expected regarding their potential to improve teaching and learning. Whereas many studies have been conducted to assess the impact of educational escape rooms on learning, results have been so far inconclusive. Several studies have reported positive learning gains, whereas others have not demonstrated such learning gains. To provide a synthesis of the existing empirical body of knowledge, we performed a meta-analysis by pooling the effect sizes across 33 published studies (5322 observations). Our results suggest that educational escape rooms are highly effective learning activities (Cohen's d = 1.4). The impact on learning was consistent across diverse fields and educational levels, regardless of team size and technology involved. Yet, educational escape rooms conducted remotely yielded smaller learning gains than those conducted face-to-face. Studies comparing educational escape rooms to other educational activities -while scarce- did not suggest a significant superiority of these activities to traditional learning activities, e.g., lectures.