• A method of extracting high-accuracy control points from ATL03 in urban areas is proposed. • Extract and classify building and ground photons in different urban areas. • The elevation and plane accuracy of control points are evaluated. • ICESat-2 ATL03 photons in urban areas can act as control points. • The land cover in urban areas has significant impact on the accuracy of control points. The ice, cloud, and land elevation satellite-2 (ICESat-2) can effectively measure global surface elevations and be used as a new data source for global elevation control. ICESat-2 has provided surface-specific data products on land/vegetation (ATL08), ice sheet (ATL06), inland water (ATL13), etc., but not on urban areas. The current research using ICESat-2 data to extract control points mostly based on products already provided, focusing on land, forest and other areas. Few studies have analyzed the feasibility of extracting control points in urban areas, as well as the position and elevation accuracy of control points. In this paper, we proposed a method for extracting high-accuracy control points from ATL03 in urban areas and evaluated the elevation and plane accuracy. First, building and ground photons were extracted and classified according to the photon shape features and neighborhood relationships, thereby addressing the insufficiency of ICESat-2 of not classifying photons in urban areas. Then, high-accuracy photons were selected from the ground photons as the laser control points considering the influence of land cover. Finally, the solid and undulating city outline was used for shape matching with DSM, thereby addressing the problem of plane accuracy evaluation in flat areas. We verified our method in two areas (Nanjing and Auckland) and showed that the method can adequately classify buildings and ground photons in different scenarios. The plane accuracy was better than 3.8 m, which lies well within ICESat-2′s mission requirement, with an elevation accuracy of better than 0.5 m, which meets the accuracy requirements of large-scale mapping, suggesting that photons in urban areas can also provide high-accuracy control data for stereo mapping.