Since Lawrence in 2001 proposed the open access (OA) citation advantage, the potential benefit of OA in relation to citation impact has been discussed in depth. The methodology to test this postulate ranges from comparing the impact factors of OA journals versus traditional ones, to comparing citations of OA versus non-OA articles published in the same non-OA journals. However, conclusions are not entirely consistent among fields, and two possible explications have been suggested in those fields where a citation advantage has been observed for OA: the early view and the selection bias postulates. In this study, a longitudinal and multidisciplinary analysis of gold OA citation advantage is developed. All research articles in all journals for all subject categories in the multidisciplinary database Web of Science are considered. A total of 1,138,392 articles-60,566 (5.3%) OA articles and 1,077,826 (94.7%) non-OA articles-published in 2009 are analysed. The citation window considered goes from 2009 to 2014, and data are aggregated for the 249 disciplines (subject categories). At journal level, we also study the evolution of journal impact factors for OA and non-OA journals in those disciplines whose OA prevalence is higher (top 36 subject categories). As the main conclusion, there is no generalizable gold OA citation advantage, neither at article nor at journal level.