Perineural invasion (PNI) is associated with poor outcomes in head and neck squamous cell carcinoma (HNSCC), yet neural invasion is seldom the outcome around which preclinical models are designed. Perineural infiltration has been observed in orthotopic tumors, but its occurrence alone does not establish whether a model can reproducibly measure nerve entry, extension along a nerve, or response to treatment. In this paper, the criteria for developing PNI-focused models are proposed, by which these outcomes can be distinguished from primary-tumor growth. Some recent studies provide two useful starting points: HO-1-u-1 tongue xenografts exhibit nerve-directed invasion, while direct infraorbital-nerve inoculation of murine oral cancer cells reproduces neural injury, nociceptive responses and impaired oral function. The former permits investigation of invasion from a mucosal tumor; the latter bypasses nerve entry and permits study of established neural disease. Model selection should therefore incorporate this experimental distinction while validation should require prespecified pathological definition, blinded and reproducible sampling, and analysis of neural involvement alongside tumor burden. Imaging and functional measurements should also be tested against histology, and the host immune context should match the mechanism or therapy under investigation. Establishing these measures would allow studies [and scientists] to determine whether an intervention prevents nerve entry, restricts neural spread or preserves function, and whether those effects extend beyond a reduction in tumor size.