OBJECTIVE: To measure changes in undiagnosed HIV among gay and bisexual men (GBM) in Melbourne. METHODS: Undiagnosed HIV was compared between GBM recruited anonymously in 2008 in gay venues only and GBM anonymously or confidentially (results delivery) recruited in 2014 at gay venues and a community festival. Surveys were completed and oral fluid specimens collected for HIV testing; positive tests among GBM reporting being HIV-negative or unknown/untested were classified as undiagnosed. Tests of proportions compared serological prevalence, undiagnosed prevalence and participant characteristics. RESULTS: HIV prevalence was 9.5% and 7.1% among 639 and 993 GBM recruited in 2008 and 2014, respectively; undiagnosed prevalence declined significantly from 31.1% to 7.1% (p<0.001). Sexual risk and undiagnosed HIV was highest among venue-recruited participants in 2014 (17.6%). Fewer diagnosed GBM participated confidentially in 2014, but this did not meaningfully influence comparative undiagnosed HIV prevalence. CONCLUSION: We provide the first estimates of changes in undiagnosed HIV in Australia, demonstrating a marked decline in undiagnosed HIV among GBM. Implications for public health: Our findings are consistent with reports of increases in HIV testing among GBM. Given sustained high HIV diagnosis rates, new testing models that encourage high frequency testing are needed to control the local HIV epidemic.