This study examined: (i) whether item nonresponse in a web survey of sexual behaviour among gay and bisexual men was lower than in an equivalent pen-and-paper survey and (ii) whether this effect was heightened for questions of a sensitive nature. Web survey respondents were recruited via Internet sites for gay men. Pen-and-paper respondents were recruited from central London gyms with a substantial gay membership. Item nonresponse was significantly lower in the web survey (1.6% vs 6.7%, p≤0.01). Item nonresponse for 78 substantially equivalent questions from each survey was highly correlated (0.82), indicating that respondents were likely to miss the same questions in each survey including sensitive questions. Our study found that item nonresponse in a web survey of sexual behaviour among gay and bisexual men was consistently lower than in an equivalent pen-and-paper survey. There was no evidence that the mode effect was heightened for questions of a sensitive nature.