“One window acceptance” is an important part of deepening the reform of “release, management and service”, and moving from “one door” to “one window” is an important measure to simplify the procedures. At present, the “one window acceptance” is still in the policy pilot stage, and studying its innovation diffusion process is of great significance to find the problems existing in the “one window acceptance” test and promote its development. Based on the self-built case base of “one window acceptance” policy documents, with the process tracing of policy experiments as the main line, this study comprehensively adopts NLP text similarity test, QGIS geospatial analysis and other methods to restore the innovation diffusion process of “one window acceptance” policy experiments from the three dimensions of “block innovation”, “learning diffusion” and “regional distribution”. Through the comparison between pilot cities and “non-pilot cities”, this study reveals two different development paths in the policy experiment, namely, official pilot and “private” exploration, tests the work effectiveness of each participant, and summarizes three innovation diffusion models-“provincial leading, independent exploration, and urban self-development”. The five key relationships of “policy orientation and pilot cities, pilot and non-pilot, superior guidance and independent development, first development and later development, edge and proximity” formed in the policy experiment of “one window acceptance” were clarified, and six innovation diffusion laws in the field of “one window acceptance” were summarized, including atypical competition, delayed influence, learning isolation, late development avoidance, edge innovation, and near-field communication. By summarizing the experience and lessons of the policy experiment, we can provide inspiration for the follow-up development.