The medical applications of carbon nanotubes (CNTs) have garnered much attention. However, evaluating the safety of CNTs remains difficult, and no consensus has been reached. Moreover, assessing the biosafety of multi-walled CNTs (MWCNTs), which can become tangled during manufacturing, is challenging because they do not readily disperse. We studied how the dispersion state of tangled MWCNTs affects their cytotoxicity, using three sonicators. Flotube 9110 (FT9110), tangled MWCNTs, were dispersed in two dispersants (fetal bovine serum and polysorbate 80) using a new type of sonicator (PR-1) and two conventional sonicators. The size and cytotoxicity of the dispersed FT9110 were measured using the BEAS-2B human bronchial epithelial cell line. The PR-1 dispersed the FT9110 to agglomerates <200 nm in diameter; FT9110 dispersed with the PR-1 did not show cytotoxicity regardless of dispersant. The other sonicators dispersed the FT9110 to particles >1000 nm in diameter, and cytotoxicity depended on the dispersant. We found that excluding cells adhered to agglomerated FT9110 before evaluating cytotoxicity can lead to false-positive results. The PR-1 sonicator dispersed tangled FT9110 to many single fibers, which showed lower cytotoxicity than conventionally-sonicated MWCNTs. We suggest that dispersion state should be accounted for when evaluating the cytotoxicity of MWCNTs.