From the WeeklyOSM article:

Les Libres Géographes has shared a methodology called OSM Skeleton, which aims to reduce the heterogeneity of OpenStreetMap data to create a coherent and locally maintainable common dataset of geographic references. The approach systematically identifies, on a national scale, missing, incomplete, inaccurate, or obsolete features among those considered fundamental: place nodes, road network, and residential areas. It uses a hybrid feature called ‘urban areas’ for analytical purposes. In particular, it generates WMS layers of alerts, instruction pages for easily correcting them in JOSM or iD, tracking statistics, and also a score for the density of POIs in towns. Séverin Ménard will present it at the upcoming SotM in Paris.

From the main article:

OpenStreetMap (OSM) relies on a wide variety of contributions in terms of mapped themes, areas covered, and levels of completeness. This flexibility is the key to its richness and success. However, it also results in a high degree of data heterogeneity, a point often cited by its critics or those who seek to downplay its value. OSM combines areas that are extremely detailed, sometimes even down to multiple floors of a single building, with others where essential information is completely missing.

This heterogeneity applies equally to data extracted from aerial imagery and to data collected on the ground. Thus, the mere presence of a categorized settlement or the road connecting it to the rest of the road network is still missing in many areas, particularly in the Global South, where the number of OSM contributors is smaller.