Urban planning relies on accurate and comprehensive spatial data to analyse existing urban configurations, design new development scenarios, and evaluate their feasibility. OpenStreetMap (OSM), the largest freely accessible and editable geographic database, is increasingly used to support such tasks. However, as a crowdsourced platform, OSM presents both strengths and vulnerabilities, particularly regarding the consistency and reliability of its data. This study assesses the accuracy of OSM data in the municipalities of Massa and Viareggio (Tuscany, Italy) – focusing in particular into commercial activities and amenities related to the 15-min city concept – with the objective of identifying and characterizing errors within the dataset. The analysis involved a comparative evaluation between point of interests (POIs) extracted from OSM and a ground-truth dataset obtained through extensive on-site surveys. Rule-based manual matching in a GIS environment was applied to align POIs entries across the two datasets and to quantify discrepancies. For each unmatched OSM entry, the type of error was classified, and results were further disaggregated by functional category. Findings reveal a general underrepresentation of POIs in the OSM dataset, along with frequent positional inaccuracies and outdated information. A limited number of redundant entries were also identified. These results underscore the inherent limitations of Volunteered Geographic Information (VGI), which can be affected by contributor discretion and uneven data coverage. Given that such data increasingly inform spatial analyses and planning decisions, their unreliability poses a critical risk to evidence-based urban design.