Substance use exacerbates psychosis, mania, depression, and poor functioning in people with first episodes of psychosis (FEP) and is associated with poor treatment outcomes, even when it does not reach the level of a formal disorder. Impaired insight and substance use are common issues that may interfere with treatment outcomes among people experiencing FEP, yet both are treatable. Improvements in these domains are associated with better outcomes. Low insight could increase risk for substance use by impairing the ability to self-appraise and assess consequences. Introspective accuracy (IA) is understudied in this area and is one way of considering self-appraisal. This study is an archival review using data collected from NAVIGATE, a coordinated specialty care program treating people with FEP. IA was operationalized as the difference between clinician and client ratings of substance use. We tested whether IA changed over one year of treatment and whether those changes occurred alongside changes in symptoms and illness self-management. No changes in IA were detected in relation to illness self-management. Changes in IA for substance use occurred midway through treatment-individuals with greater symptom remission had more overconfident IA. Prior research on insight has shown a paradox where greater insight accompanies more symptoms. However, past research has also shown a relationship between IA and functional outcomes, like illness self-management, and that overconfidence in one domain can positively bias clinician ratings in another. Our findings suggest either a positive bias for ratings associated with overconfident IA or an insight paradox type effect.