Using language to distance oneself from negative stimuli (e.g., by reducing use of the word “I” and present-tense verbs) is associated with effective emotion regulation. Given that internalizing disorders like anxiety and depression are characterized by maladaptive emotion regulation, stronger linguistic distance may be both a diagnostic marker of lower internalizing symptoms and a prognostic indicator of treatment progress. Here, we tested these hypotheses in a large corpus of naturalistic psychotherapeutic exchanges between clients and their therapists (> 1.2 million messages from 6,229 clients). In both exploratory (N=3,729) and validation (N=2,500) datasets, we found that clients’ internalizing symptoms decreased over therapy, that client linguistic distance increased over therapy, and that internalizing symptoms tracked fluctuations in linguistic distance both within- and between-individuals. In other words, clients shifted from discussing themselves and the present moment to discussing other people and timepoints over treatment, and this was related to symptom improvements. However, effect sizes for linguistic results were small, and we failed to find consistent evidence that linguistic distance statistically mediated changes in symptoms over time. Treatment efficacy also was related to therapist linguistic distance in some—but not all—analyses, suggesting that changing therapist language could improve outcomes. Finally, clustering analyses revealed that data-driven groups of clients defined based solely on their linguistic distance differed in both their symptom severity and treatment outcomes. Together, these findings provide replicable evidence that linguistic distance is a marker of internalizing symptom severity and treatment progress in real-world therapeutic interactions.