The rich information hidden in misspoken discourse.

This commentary focuses on the use of artificial intelligence (AI) natural language processing tools to classify primary progressive aphasia (PPA) variants based on connected speech analysis. Rezaii et al. applied large language models to speech samples from individuals with PPA, successfully identifying three main variants—non-fluent agrammatic, logopenic, and semantic—that correspond to distinct patterns of cortical atrophy and clinical diagnoses. Each PPA variant is linked to specific underlying neuropathologies and divergent clinical progressions, making accurate classification important for prognosis and potential treatment strategies. The study also highlights linguistic features, such as verb frequency usage, that differentiate the variants, providing insights into how different brain regions contribute to language processing. This work demonstrates the potential of AI to enhance understanding of neurodegenerative language disorders through detailed speech analysis.