Objectives: To examine changes in the sexual behavior of London gay men between 1998 and 2008. Methods: Gay men using London gyms were surveyed annually between 1998 and 2005, and again in 2008 (n = 6064; range, 482-834 per year). Information was collected on human immunodeficiency virus (HIV) status of the respondent, unprotected anal intercourse (UAI) in the previous 3 months, type (main or casual) and HIV status of partner for UAI. Nonconcordant UAI (ncUAI) was defined as UAI with a partner of unknown or discordant HIV status. Concordant UAI (cUAI) was defined as UAI with a partner of the same HIV status ("serosorting"). Results: Between 1998 and 2008, the percentage of men reporting UAI increased from 24.3% to 36.6% (P = 0.07). This overall increase concealed important differences between nonconcordant and concordant UAI. While the percentage of men engaging in cUAI increased steadily between 1998 and 2008 (9.8%, 20.8%; P = 0.01), the percentage reporting ncUAI increased between 1998 and 2001 (14.5%, 23.7%; P < 0.001), decreased between 2001 and 2005 (23.7%, 15.6%; P < 0.001), and then leveled off between 2005 and 2008 (15.6%, 15.7%; P = 0.2). However, the percentage of men reporting ncUAI with a main partner increased between 2005 and 2008 for HIV-positive men (2.5%, 8.1%; P < 0.05) and HIV negative men (2.1%, 5.5%; P = 0.06). While the percentage of HIV negative men who reported cUAI with a main partner (i.e., serosorting) increased between 1998 and 2008 (12.4%, 21.1%; P < 0.05), less than half established seroconcordance by testing together. Conclusions: The patterns of sexual behavior among London's gay men between 1998 and 2008 were dynamic and complex. Our data suggest that HIV risk with a main partner and HIV testing among couples should be given greater priority by health promotion programmes.