Introduction. Patients with a fluent type of aphasia have semantically poor discourse, which is often uninformative for the interlocutor, having consequences for their daily life. One of the treatment methods used in clinical practice is the semantic feature analysis (SFA). The aim of this study was to determine the impact of this treatment method on the content of discourse in patients with fluent aphasia. Methods. The sample consisted of 30 subjects with fluent aphasia, namely: Wernicke's aphasia (9 subjects), conductive aphasia (6 subjects), transcortical sensory aphasia (8 subjects), and anomic aphasia (7 subjects). To assess the formation of narrative discourse, the picture subtest "Cake theft" from the Boston diagnostic test for aphasia was used. The content of the discourse was assessed by two experienced speech therapists - aphasiologists. Results. After the treatment using the semantic feature analysis, an increase in the number of content words was observed in all patients (H = 22.53, df = 3, p < .001; H = 23.42, df = 3, p < .001; H = 23.10, df = 3, p < .001). Patients with Wernicke's and transcortical sensory aphasia had most impairment in the content/informativeness of discourse. Uninformative discourse was observed in 5 (33.33%) patients with Wernicke's and 4 (26.67%) patients with transcortical sensory aphasia. The type and severity of aphasia were shown to affect the informativeness of discourse. Conclusion. The semantic feature analysis treatment method leads to a significant improvement of narrative discourse in patients with milder forms of aphasia, such as anomic and conductive aphasia.