Persistence length of (bio)macromolecules plays an increasingly important role in macromolecular science, especially in emerging fields like macromolecular biophysics. It is shown that using established approximate relations for the estimation of persistence length for semiflexible macromolecules leads to differences between different methods or even to a modified behavior, especially at high chain stiffness. The approximate estimate obtained from the average bond angle (B) performs better than the estimate from the decay of bond orientation correlation (C) in chain but close to the coil-to-rod transition starts to fail. This approximate estimate is close to the estimate from the definition of persistence (D) (with minimum approximations) not only in the theta state or the random coil but also in the good solvent. The value of persistence from the decay of bond orientation correlation along the macromolecule is underestimated when compared to the exact value. The persistence obtained from the worm-like chain model (W) is close to the exact result in the coil regime (also in the good solvent) but near at and above the transition produces the largest values of all approximations. In systems with interactions this model starts to underestimate the persistence on decreasing temperature. We advocate using the estimate (D) of persistence length generally since it avoids limitations mentioned above and at the same time does not pose additional requirements on the evaluation itself. (C) 2004 Elsevier Ltd. All rights reserved.