The paper presents an ultrafast bootstrap approximation method (UFBoot) for phylogenetic analysis. UFBoot is designed to compute the support of phylogenetic groups in maximum likelihood (ML) based trees. It combines the resampling estimated log-likelihood method with a simple but effective collection scheme of candidate trees. A stopping rule is also proposed to assess the convergence of branch support values and automatically determine when to stop collecting candidate trees. UFBoot achieves a median speed up of 3.1 to 10.2 compared with RAxML RBS for real DNA and amino acid alignments, respectively. The method is robust against moderate model violations and provides relatively unbiased support values compared to the conservative standard bootstrap. An efficient and easy-to-use software (IQ-TREE) is provided for performing UFBoot analysis with ML tree inference. The study compares UFBoot with other bootstrap approaches in terms of accuracy and computational time. UFBoot is found to be almost unbiased compared to other methods for both simulations, with support values closely reflecting the probabilities of the split being correct. UFBoot is also more efficient than other methods, with computational times ranging from 3.1 to 10.2 times faster for DNA and amino acid alignments, respectively. The method is robust against moderate model violations and provides a more direct interpretation of the bootstrap support. The study also highlights the importance of assessing phylogenetic signals in the data before carrying out an expensive bootstrap analysis. UFBoot is a time-saving option compared with other bootstrap inference tools. The method is implemented in the IQ-TREE software package, which allows users to reconstruct the ML tree, bootstrap trees, and consensus tree within one single run. The study concludes that UFBoot is a fast and accurate method for phylogenetic analysis.The paper presents an ultrafast bootstrap approximation method (UFBoot) for phylogenetic analysis. UFBoot is designed to compute the support of phylogenetic groups in maximum likelihood (ML) based trees. It combines the resampling estimated log-likelihood method with a simple but effective collection scheme of candidate trees. A stopping rule is also proposed to assess the convergence of branch support values and automatically determine when to stop collecting candidate trees. UFBoot achieves a median speed up of 3.1 to 10.2 compared with RAxML RBS for real DNA and amino acid alignments, respectively. The method is robust against moderate model violations and provides relatively unbiased support values compared to the conservative standard bootstrap. An efficient and easy-to-use software (IQ-TREE) is provided for performing UFBoot analysis with ML tree inference. The study compares UFBoot with other bootstrap approaches in terms of accuracy and computational time. UFBoot is found to be almost unbiased compared to other methods for both simulations, with support values closely reflecting the probabilities of the split being correct. UFBoot is also more efficient than other methods, with computational times ranging from 3.1 to 10.2 times faster for DNA and amino acid alignments, respectively. The method is robust against moderate model violations and provides a more direct interpretation of the bootstrap support. The study also highlights the importance of assessing phylogenetic signals in the data before carrying out an expensive bootstrap analysis. UFBoot is a time-saving option compared with other bootstrap inference tools. The method is implemented in the IQ-TREE software package, which allows users to reconstruct the ML tree, bootstrap trees, and consensus tree within one single run. The study concludes that UFBoot is a fast and accurate method for phylogenetic analysis.