Recent reports of CRISPR- (clustered regularly interspaced short palindromic repeats)/Cas9 (CRISPR-associated protein 9) mediated heritable mutagenesis in plants highlight the need for accuracy of the mutagenesis directed by this system. Off-target mutations are an important issue when considering functional gene analysis, as well as the molecular breeding of crop plants with large genome size, i.e. with many duplicated genes, and where the whole-genome sequence is still lacking. In mammals, off-target mutations can be suppressed by using Cas9 paired nickases together with paired guide RNAs (gRNAs). However, the performance of Cas9 paired nickases has not yet been fully assessed in plants. Here, we analyzed on- and off-target mutation frequency in rice calli and regenerated plants using Cas9 nuclease or Cas9 nickase with paired gRNAs. When Cas9 paired nickases were used, off-target mutations were fully suppressed in rice calli and regenerated plants. However, on-target mutation frequency also decreased compared with that induced by the Cas9 paired nucleases system. Since the gRNA sequence determines specific binding of Cas9 protein–gRNA ribonucleoproteins at the targeted sequence, the on-target mutation frequency of Cas9 paired nickases depends on the design of paired gRNAs. Our results suggest that a combination of gRNAs that can induce mutations at high efficiency with Cas9 nuclease should be used together with Cas9 nickase. Furthermore, we confirmed that a combination of gRNAs containing a one nucleotide (1 nt) mismatch toward the target sequence could not induce mutations when expressed with Cas9 nickase. Our results clearly show the effectiveness of Cas9 paired nickases in delivering on-target specific mutations.