Online data collection could advantageously replace paper-and-pencil questionnaires in epidemiological studies by reducing the logistic burden, the cost and the duration of data processing. However, there is a need for studies comparing these new instruments to traditional ones. Our objective was to compare the web-based version of the NutriNet-Santé self-administered anthropometric questionnaire to the paper-based version. The questionnaire included 17 questions divided into subquestions (55 variables in all) dealing with height, weight, hip and waist circumferences, weight history, restrictive diet and weight self-perception. Both versions of the questionnaire were filled out by 147 volunteers (paper version first, N = 76, or web-based version first, N = 71) participating in the SU.VI.MAX ("Supplémentation en VItamines Minéraux et AntioXydants") cohort (age-range: 49-75 years; men: 46.3%). At the end of the test, subjects filled in a "satisfaction" questionnaire giving their opinions and feelings about each version. Agreement was assessed by intraclass correlation coefficients (ICCs) and kappas. We also quantified the number of errors inherent in the paper version. Agreement between the two versions was high. ICCs ranged from 0.86 to 1.00. Kappas ranged from 0.69 to 1.00 for comparable variables. A total of 82 data entry mistakes (1.5% of total entries), 60 missing values (1.1%), 57 inconsistent values (1.1%) and 3 abnormal values (0.1%) were counted in the paper version (non-existent in the web-based version due to integrated controls). The web-based version was preferred by 92.2% of users. In conclusion, the quality of information provided by the web-based anthropometric questionnaire used in the NutriNet-Santé Study was equal to, or better than, that of the paper version, with substantial logistic and cost advantages.