This systematic review investigates the role of virtual reality (VR)-based multisensory cognitive training in cognitive function, executive function and wayfinding ability among people diagnosed with mild cognitive impairment (MCI) and Alzheimer's disease (AD). The review was carried out using PRISMA guidelines. PubMed, Scopus, Embase, and Google Scholar were searched up from inception to February 2025 using terms related to MCI, AD, VR, and cognitive functions. Studies were included if they involved participants with MCI or early AD, used VR-based training, collected baseline data, and reported cognitive outcomes. Nine studies with MCI were included, but no eligible studies focused on AD. Seven out of nine eligible studies in MCI reported significant improvements in global cognitive function (MoCA, CERAD-K, MMSE). Some studies showed improvements in executive function (EXIT-25, TMT-A/B, and SCWT), while others found no significant differences. One study reported improved depression/mental status (GDS, MOSES, QoL-AD). Just one study reported improvement in functional ability (IADL). One study reported enhanced cognition and reduced discomfort (SSQ). VR programs were generally well-tolerated, with no significant adverse events reported. VR shows promise for improving cognitive function in MCI. VR also showed potential benefits in executive function and psychological outcomes like depression and quality of life, though consistency varied.