OpenStreetMap (OSM) enters its third decade amid pronounced change and instability. Corporate involvement continues to evolve. Since it was last documented in 2019 [1], corporate mapping - instrumental in expanding OSM beyond its early European core [2] - has declined [3] as the Overture Maps Foundation has emerged as a parallel venue for producing OSM-derived map products. COVID-19 caused further disruption: between 2020 and 2023, only one global State of the Map was held in person; contributor engagement and co-editing changed [4]; and the project gained visibility, contributing to record activity in May 2020 [5]. Geopolitical conflicts have likewise exposed OSM's close ties to events on the ground. Politically motivated vandalism during the war in Gaza prompted new rate limits and revealed tensions among openness, governance, and resilience [6], while bombardment and shifting front lines may have reshaped building contributions in Ukraine [7]. Meanwhile, OSM continues to expand into new communities and domains [2], even as large language models and other AI systems consume its data and infrastructure at unprecedented scale, raising questions about attribution, reciprocity, and the sustainability of a volunteer commons [8].
Academic research usually proceeds more slowly than real-world events because funding, research programs, and publication take time. Conference presentations, however, can reveal emerging trends because they accommodate ongoing and unpublished work. Indeed, topics presented in State of the Map talks across all tracks appeared in journal publications an average of 2.8 years later [9]. The abstracts in OSM Science 2026 can therefore be read as a timely, if necessarily provisional, scientific response to a project being reshaped by new actors, crises, technologies, and uses.
Several topics outlined above recur in this edition of OSM Science 2026. Rihani et al. compare events on the front line of the war in Ukraine with patterns of building edits [7]; Boettcher examines digital-governance problems arising from AI-driven extraction of OSM data [8]; and Grinberger et al. analyze the long-term effects of corporate mapping on OSM's data landscape in Southeast Asia [10].
These contributions sit alongside studies of 'classic' OSM Science themes. Despite the project's maturity, data quality continues to attract substantial attention. A substantial group of papers addresses it directly, while several related contributions combine quality assessment with tools or applications. Levin et al. propose an intrinsic completeness indicator that uses the geometric attributes of mapped entities to identify potentially incomplete regions [11]. Bourbonnais et al. quantify the errors introduced when POI centroids are used as proxies for actual entrances [12], while Tauscher and Enkhtamir estimate the amount of potentially mappable publicly accessible indoor space and provide a basis for calculating the completeness of indoor mapping [13]. Bolognani and Herfort conduct a relation-aware audit of the Dutch cycling network [14]. Flemming's Public Transport Stop Analysis project, which offers a technical contribution to enable rapid, planet-scale calculations, visualizes mapping conventions and flags inconsistent or potentially erroneous data [15]. Finally, Mazagol identifies meaningful gaps and classification errors in the mapping of mining activities relative to scientific and national inventories [16], while Sarretta et al. document sparse OSM coverage, very low adoption of dedicated tags, and missing condition and maintenance attributes for Alpine torrent-control structures, and propose steps toward a more complete interoperable inventory [17].
These studies focused on data are complemented by efforts to make OSM data and contributions more accessible and interpretable. Some focus on understanding edits: Bernard et al.'s GeoChange Ontology, processing pipeline, knowledge graph, and web application classify transitions to distinguish real-world change from correction, refinement, and modeling decisions [18], while Rieger et al.'s OSM Road Monitor compares road edits with masks derived from recent remote-sensing imagery to flag possible inaccuracies [19]. Other contributions take a governance-oriented approach. Liebel examines the Humanitarian OpenStreetMap Team (HOT) Tasking Manager as a sociotechnical interface that can both address and reproduce asymmetries in humanitarian mapping [20], while Cámara-Menoyo and Monteath analyze diversity and exclusion in OSM's formal tagging-proposals process [21]. Further studies support mapping directly. StreetMeasure combines semantic segmentation, monocular depth estimation, crowdsourced 360-degree imagery, and limited field measurements to recommend measurement tags [22]. Finally, two cartographic tools make OSM-derived outputs more accessible. Bourcier and Touya present the use of CartAGen, an open-source Python library containing more than 80 generalization algorithms and offering a QGIS processing interface, on OSM data with the goal of designing a high-quality pan-scalar map [23]. Amanuel studies the rapid uptake of Terraink, a service that converts OSM data into printable city posters without requiring GIS or programming expertise, using its release to examine non-expert demand and the infrastructural effects of increased consumption [24].
Despite persistent data-quality issues and the continuing need for new tools, OSM remains widely used, as demonstrated by the many domain-specific studies in this edition 2026. Most concern transport and mobility: Rao proposes reconstructing high-resolution lane-level road data from OSM centerlines [25]; Szeliga et al. describe using OSM-derived variables to model operating speed through spatial regression [26]; Votz describes inferring applicable speed limits on an offline smartphone and comparing them with measured vehicle speed [27]; Klinkhardt et al. discuss deriving attractiveness scores for destination-choice models [28]; and Jeansoulin and Gambette deliver enrichment of historical railway archives [29]. A second group of studies address welfare, health, and humanitarian concerns: Ihantamalala et al. model access to health services in rural Madagascar [30], Loizeau et al. assess the health potential of urban green spaces [31], and van der Heijden describes detecting slums at the individual-building level [32]. A final, unusual application proposed by Struś et al. combines OSM-based route planning and urban-context information with community mapping, guided soundwalks, and building-level color observations to create multisensory urban experiences [33].
Taken together, these contributions span domain applications, data-quality assessment, analyses of contribution dynamics and editing behavior, the development of tools that support OSM contributions, and questions of governance and participation. These themes echo those identified by the Scientific Committee of the 2019 State of the Map Academic Track, the predecessor of OSM Science [34]. This apparent continuity is also visible across the full set of Academic Track and OSM Science abstracts. Using a large-language-model-based process with human oversight and intervention, we classified all 122 non-editorial abstracts published between 2019 and 2026, allowing each abstract to be assigned to more than one category. The results (see Figure 1) show that every category is meaningfully represented in almost every year, although several patterns emerge. Data quality is the most stable topic, consistently accounting for about one-fifth of contributions. The development of tools that support OSM, while showing more variability, has also remained relatively stable since 2021. By contrast, attention to contribution dynamics and editing behavior appears to have declined as studies of governance and participation have become more common (even if somewhat declining recently). Domain applications have also increased and now form a consistently prominent category.