UFBoot2 is an improved version of the ultrafast bootstrap (UFBoot) method for phylogenetic analysis. UFBoot was developed to reduce computational time while providing more unbiased branch support than the standard bootstrap (SBS). UFBoot2 further accelerates UFBoot and reduces the risk of overestimating branch support due to polytomies or severe model violations. It also provides suitable bootstrap resampling strategies for phylogenomic data. UFBoot2 is significantly faster than SBS and RAxML rapid bootstrap. It is 778 times faster than SBS and 8.4 times faster than RAxML rapid bootstrap on tested datasets. UFBoot2 is implemented in the IQ-TREE software package version 1.6 and is freely available at http://www.iqtree.org. The key improvements in UFBoot2 include accelerating the likelihood computation through modifications to the pruning algorithm and the use of SIMD (single instruction, multiple data) features. Additionally, UFBoot2 corrects for polytomies by selecting a tree with a RELL score less than a threshold from the trees encountered during the search, rather than the one with the highest RELL score. This reduces the risk of overestimating branch support for short branches. UFBoot2 also addresses the impact of model violations by providing an option to further optimize the best RELL-trees using a hill-climbing nearest-neighbor interchange (NNI) search. This step, called UFBoot2 + NNI, helps to reduce the overestimation of branch supports in the presence of severe model violations. UFBoot2 also implements several resampling strategies for phylogenomic data, including resampling MSA-sites within partitions, resampling genes, and resampling genes and then resampling MSA-sites within each gene. These strategies allow for different numbers of sites in the bootstrap MSAs and can lead to different support values for branches. UFBoot2 is a fast and accurate alternative to other bootstrap methods, particularly under no or mild model violations. It provides unbiased bootstrap support as suggested for UFBoot. Users are advised to apply model violation detection methods before bootstrap analyses. UFBoot2 + NNI should be applied if severe model violations are present in the data set.UFBoot2 is an improved version of the ultrafast bootstrap (UFBoot) method for phylogenetic analysis. UFBoot was developed to reduce computational time while providing more unbiased branch support than the standard bootstrap (SBS). UFBoot2 further accelerates UFBoot and reduces the risk of overestimating branch support due to polytomies or severe model violations. It also provides suitable bootstrap resampling strategies for phylogenomic data. UFBoot2 is significantly faster than SBS and RAxML rapid bootstrap. It is 778 times faster than SBS and 8.4 times faster than RAxML rapid bootstrap on tested datasets. UFBoot2 is implemented in the IQ-TREE software package version 1.6 and is freely available at http://www.iqtree.org. The key improvements in UFBoot2 include accelerating the likelihood computation through modifications to the pruning algorithm and the use of SIMD (single instruction, multiple data) features. Additionally, UFBoot2 corrects for polytomies by selecting a tree with a RELL score less than a threshold from the trees encountered during the search, rather than the one with the highest RELL score. This reduces the risk of overestimating branch support for short branches. UFBoot2 also addresses the impact of model violations by providing an option to further optimize the best RELL-trees using a hill-climbing nearest-neighbor interchange (NNI) search. This step, called UFBoot2 + NNI, helps to reduce the overestimation of branch supports in the presence of severe model violations. UFBoot2 also implements several resampling strategies for phylogenomic data, including resampling MSA-sites within partitions, resampling genes, and resampling genes and then resampling MSA-sites within each gene. These strategies allow for different numbers of sites in the bootstrap MSAs and can lead to different support values for branches. UFBoot2 is a fast and accurate alternative to other bootstrap methods, particularly under no or mild model violations. It provides unbiased bootstrap support as suggested for UFBoot. Users are advised to apply model violation detection methods before bootstrap analyses. UFBoot2 + NNI should be applied if severe model violations are present in the data set.