This study aimed to investigate cortical activation and functional connectivity during Stroop task performance in Parkinson's disease (PD) using functional near-infrared spectroscopy (fNIRS). Forty-five individuals with PD and fourteen healthy controls completed neuropsychological assessments and underwent fNIRS scanning while performing the Stroop task. PD participants were categorized into normal cognition (PD-NC, n = 6), mild cognitive impairment (PD-MCI, n = 22), and dementia (PDD, n = 17) groups. Z scores were calculated across cognitive domains, including attention, working memory, executive function, language, memory, and visuospatial function. During the Stroop task, significant hypoactivation in the dorsolateral prefrontal cortex (DLPFC), primary motor cortex (M1), and premotor cortex (PMC) were observed in the PD-MCI group, while PDD patients showed increased activation in the medial prefrontal cortex (mPFC), orbitofrontal cortex (OFC), and dorsolateral prefrontal cortex (DLPFC). Increased activation in DLPFC was significantly correlated with poorer executive function outcomes. Functional connectivity analysis revealed that both PD-NC and PD-MCI groups had significantly enhanced interhemispheric connectivity compared to healthy controls, with pronounced interhemispheric connectivity in PD-MCI. In contrast, the PDD group exhibited reduced connectivity among the premotor cortex (PMC), ventrolateral prefrontal cortex (VLPFC), and orbitofrontal cortex (OFC), compared to the PD-MCI group. While PD-MCI patients showed reduced cortical activation relative to PDD, they exhibited extensive connectivity across cortical regions, suggesting an expanded cortical network as compensation. In PDD, the mPFC, left OFC, and left DLPFC displayed the highest cortical activation and alongside reduced functional connectivity, which may reflect widespread atrophy across multiple brain regions. These findings highlight fNIRS as a potential tool for characterizing cognitive impairment stages in PD.