Objective The Diagnostic and Statistical Manual of Mental Disorders 5 suggests the 5th age‐adjusted body mass index (BMI) percentile as the numeric cut‐off for anorexia nervosa (AN) in children and adolescents. We aimed to investigate the degree to which the 5th age‐adjusted percentile as the numeric cut‐off for AN in youths reflects the clinical population of patients accepted for treatment. Method From a specialized eating disorder clinic, 305 patients with AN below 18 years of age were grouped according to age‐adjusted BMI percentiles [below the 5th (low), above the 10th (high), and between the 5th and the 10th (medium)]. The distribution of eating disorder diagnoses and severity measured by the Eating Disorder Examination was compared. Results Full‐syndrome anorexia nervosa (F.50.0) was found in 182 (59.5%) patients and atypical anorexia nervosa (F.50.1) in 123 patients (40.5%). The number of patients in the low, medium, and high BMI percentile groups was 189 (62.0%), 34 (11.1%), and 82 (26.9%), respectively. Patients in the low BMI group differed from patients in the medium BMI group by a lower frequency of vomiting. The high BMI group presented with more weight and shape concern than the lower BMI group. Age was not a confounder of these associations. Discussion We question the applicability of the 5th BMI percentile as a substantiated cut‐off for the weight criterion in anorexia nervosa in youths and argue that the cut‐off should not be ascribed great clinical importance as this may hinder early detection of illness and initiation of treatment.