Abstract: Objectives: Real-time Global Ionospheric Maps (GIMs) can provide prior ionospheric constraints for precise point positioning (PPP). However, missing or unreliable root mean square (RMS) information may lead to inappropriate stochastic weighting of ionospheric pseudo-observations and thus compromise positioning performance. This study aims to evaluate the reliability of real-time GIM RMS information and to generate regional RMS maps for ionosphere-constrained single-frequency BeiDou Navigation Satellite System (BDS) PPP. Methods: High-precision slant total electron content (STEC) was derived from BDS dual-frequency undifferenced and uncombined PPP with ambiguity resolution. After the combined satellite-and-receiver differential code bias was removed, the derived STEC was used as a reference to construct slant-path errors for real-time GIM corrections. A regional RMS-generation method was developed by combining a two-hour sliding window, double-sided median absolute deviation (MAD) robust screening, and inverse-distance weighting. The generated RMS maps were assessed using RMS bounding percentages (RMSBPs) and normalized residual statistics. Experiments were conducted using 51 days of real-time GIM products and GNSS observations in the Asian low- and mid-latitude region. Five real-time GIM products provided by CAS, CNES, NRCan, UPC, and WHU were evaluated. Simulated real-time single-frequency BDS PPP tests were further performed at seven independent stations using CNES real-time orbit, clock, and GIM products. The proposed time-varying regional-RMS constraint (IONR) was compared with an unconstrained solution (ION0) and fixed vertical-TEC uncertainty constraints of 2 TECU (ION2) and 8 TECU (ION8). Results: After double-sided MAD screening, the mean 1-, 2-, and 3-RMSBPs for the five real-time GIM products increased from 63.1%, 96.3%, and 99.2% to 70.2%, 97.9%, and 99.7%, respectively. The mean standard deviation of normalized residuals increased from 0.85 to 0.94, approaching the ideal value of 1. Meanwhile, the mean and median slant-path RMS values decreased by approximately 13.9% and 14.6%, respectively, indicating improved consistency between the estimated RMS and the actual GIM errors. In kinematic PPP, IONR reduced the combined horizontal and vertical positioning RMS from 0.81 m and 0.80 m under ION0 to 0.53 m and 0.63 m, corresponding to improvements of 34.6% and 21.3%, respectively. In static PPP, the mean horizontal and vertical convergence times decreased from 126.0 min and 67.9 min to 96.3 min and 55.8 min, representing reductions of 23.6% and 17.8%, respectively. Overall, IONR outperformed the fixed-uncertainty constraint schemes in both positioning accuracy and convergence performance. Conclusions: The proposed method provides statistically more consistent RMS uncertainty information for real-time GIM corrections and enables adaptive ionospheric weighting in single-frequency BDS PPP. Under calm to weakly disturbed ionospheric conditions in the Asian low- and mid-latitude region, the generated regional RMS maps improve kinematic positioning accuracy and accelerate static PPP convergence. Further validation using longer observation periods, additional regions, stronger ionospheric disturbances, and multi-GNSS observations is required to assess the broader applicability of the method.