Background: With disease progression, patients with Parkinson's disease (PD) may have chronic visual hallucinations (VH). The mechanisms behind this invalidating non-motor symptom remain largely unknown, namely because it is extremely difficult to capture hallucination events. This study aimed to describe the patterns of brain functional changes when VH occur in PD patients. Methods: Nine PD patients were enrolled because of their frequent and chronic VH (> 10/day). Patients with severe cognitive decline (MMSE<18) were excluded. Patients were scanned during ON/OFF hallucinatory states and resting-state functional imaging (rs-fMRI) was performed. Data were analyzed in reference to the two-step method, which consists in: (i) a data-driven analysis of per-hallucinatory fMRI data, and (ii) selection of the components of interest based on a post-fMRI interview. Results: The phenomenology of VH ranged from visual spots to distorting faces. First, at the individual level, several VH-related components of interest were identified and integrated in a second-level analysis. Using a random-effects self-organizing-group ICA, we evidenced increased connectivity in visual networks concomitant to VH, encompassing V2, V3 and the fusiform gyri bilaterally. Interestingly, the stability of the default-mode network (DMN) was found positively correlated with VH severity (spearman's rho = 0.77, p = 0.05). Conclusion: By using a method that does not need online self-report, we showed that VH in PD patients were associated with functional changes in associative visual cortices, possibly linked with strengthened stability of resting-state networks.