Spontaneous speech contains a wealth of information that reflects personal characteristics of the speaker, such as mood, motivation, intelligence, arousal, and variability in word use. Recent advances in Natural Language Processing (NLP) have paved the way for systematic recording and near real-time analysis of quantifiable properties of spoken language. NLP can reliably provide variables relevant to various aspects of brain functioning within seconds, while the cost and effort of speech recording is negligible. In this thesis, we investigated the use of state-of-the-art NLP models to support the diagnosis of psychotic disorders (e.g., schizophrenia). Psychiatric diagnoses are currently not reliable as no objective quantitative biomarkers are available. This is a serious social problem, because incorrect diagnoses lead to over- and under-treatment. NLP analyzes of spontaneous speech provide reproducible quantitative assessment. In this thesis, we have shown that acoustic, semantic and grammatical aspects of language can be quantified and used as a marker for psychotic disorders. Based on these analyses, we can say with ~85% certainty whether someone has a psychosis or not. In addition, we have shown that computational language analyzes provide clinically relevant insights in the study of auditory verbal hallucinations. In the future, these analyzes may be used to detect a relapse in psychosis earlier, so that you can see a psychosis coming before people become seriously ill.