Different studies of same-sex couples based on the Generations and Gender Survey (GSS) have been published in recent years. However, in this survey, people in gay and lesbian couples are not identified via a direct question but by comparing two separate variables: the sex of the respondent and that of the partner. But the identification of same-sex couples by comparing the sex of both partners is very unreliable. The few sex-coding errors of one of the partners in heterosexual couples generate a very high proportion of “false” same-sex couples among all couples considered to be of the same sex. The aim of this paper is to demonstrate the problematic nature of the indicator in order to avoid inappropriate use of GGS data (and other surveys more generally) and to discuss the reliability of analyses of same-sex couples made with these data.