Introduction:　Despite　being　a　widely　used　screening　questionnaire,　there　is　no　consensus　on　the　most　appropriate　measurement　model　for　the　Alcohol　Use　Disorders　Identification　Test　(AUDIT).　Furthermore,　there　have　been　limited　studies　on　its　measurement　invariance　across　cross-cultural　subgroups,　genders,　and　sexual　orientations.　Aims:　The　present　study　aimed　to　examine　the　fit　of　different　measurement　models　for　the　AUDIT　and　its　measurement　invariance　across　a　wide　range　of　subgroups　by　country,　language,　gender,　and　sexual　orientation.　Methods:　Responses　concerning　past-year　alcohol　use　from　the　participants　of　the　cross-sectional　International　Sex　Survey　were　considered　(N　=　62,943;　Mage:　32.73;　SD　=　12.59).　Confirmatory　factor　analysis,　as　well　as　measurement　invariance　tests　were　performed　for　21　countries,　14　languages,　three　genders,　and　four　sexual　orientation　subgroups　that　met　the　minimum　sample　size　requirement　for　inclusion　in　these　analyses.　Results:　A　two-factor　model　with　factors　describing　＇alcohol　use＇　(items　1-3)　and　＇alcohol　problems＇　(items　4-10)　showed　the　best　model　fit　across　countries,　languages,　genders,　and　sexual　orientations.　For　the　former　two,　scalar　and　latent　mean　levels　of　invariance　were　reached　considering　different　criteria.　For　gender　and　sexual　orientation,　a　latent　mean　level　of　invariance　was　reached.　Conclusions:　In　line　with　the　two-factor　model,　the　calculation　of　separate　alcohol-use　and　alcohol-problem　scores　is　recommended　when　using　the　AUDIT.　The　high　levels　of　measurement　invariance　achieved　for　the　AUDIT　support　its　use　in　cross-cultural　research,　capable　also　of　meaningful　comparisons　among　genders　and　sexual　orientations.