Objective Early response to eating disorders treatment is thought to predict a later favourable outcome. A systematic review of the literature and meta‐analyses examined the robustness of this concept. Method The criteria used across studies to define early response were summarised following the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses guidelines. Diagnostic Test Accuracy methodology was used to estimate the size of the effect. Results Findings from 24 studies were synthesized and data from 14 studies were included in the meta‐analysis. In Anorexia Nervosa, the odds ratio of early response predicting remission was 4.85(95%CI: 2.94–8.01) and the summary Area Under the Curve (AUC) = .77. In Bulimia Nervosa, the odds ratio was 2.75(95%CI:1.24–6.09) and AUC = .67. For Binge Eating Disorder, the odds ratio was 5.01(95%CI: 3.38–7.42) and AUC = .71. Conclusion Early behaviour change accurately predicts later symptom remission for Anorexia Nervosa and Binge Eating Disorder but there is less predictive accuracy for Bulimia Nervosa. Copyright © 2016 John Wiley & Sons, Ltd and Eating Disorders Association.