Objective Cultural ecosystem service (CES) refer to the intangible benefits that humans obtain from ecosystems, such as spiritual satisfaction, cognitive development, and aesthetic experiences. The assessment of their value is of great significance for revealing the mechanisms of ecosystem benefits and enhancing human well-being. This research aims to systematically analyze the progress of CES assessment driven by multi-source big data, explore cutting-edge trends, and identify future directions.

Methods The search query for the Web of Science core database was TS = (“ecosystem*” AND “cultural service*”) AND (TS = valuation OR TS = evaluate OR TS = evaluation OR TS = assessment OR TS = quantification). The search query for the China National Knowledge Infrastructure (CNKI) database was SU%= “value assessment” and SU%= “evaluation” and SU% = “quantitative research” and SU% = “ecosystem cultural services”. We conducted separate searches for foreign-language and Chinese-language literature, manually screened the literature that applied multi-source big data for CES assessment, and ultimately obtained 273 foreign-language articles and 246 Chinese-language articles from 2000 to 2024. The study organized research findings across four dimensions: big data types, CES value types, assessment objects, and assessment methods. It also discussed current research opportunities, challenges, and future trends. Based on the literature review results, this study systematically constructed a CES assessment framework based on multi-source big data, with a core workflow comprising “assessment object-CES value type-big data collection-assessment method” and four core modules.

Results 1) The CES assessment paradigm is shifting from traditional economic accounting to intelligent assessment. Statistics show that approximately 70% of CES assessment studies have achieved paradigm innovations through the application of multi-source big data, primarily manifested in four aspects: expansion of CES value types, refinement of assessment objects, and innovation in assessment methods. 2) With the widespread application of big data, the data foundation for CES assessment has broken through traditional limitations, forming a diversified landscape combining government-published data (such as ecological environment data, population and economic data, etc.) with user-generated data (such as social media data, point of interest (POI) data, location-based communication data, etc.). Research progress in CES assessment system has shown a progressive trend: from early reliance on government-disclosed data, to the expansion of user-generated data, and then to multi-modal data. This trend has significantly improved the accuracy, spatio-temporal coverage, and scenario applicability of assessment research. A deeper change lies in the fact that the diversification of data sources is driving a shift in the CES assessment paradigm from “supply-driven” to “supply-demand coordination”. 3) As the application and adaptability of multi-source big data have improved, the development of assessment objects has shown a trend toward focusing and refining the scope of research, shifting from early regional-scale natural ecosystems (farmland, forests, marine ecosystems) to urban built environments, with a focus on densely populated urban ecosystems (urban green spaces, urban parks, green infrastructure). From 2020 to 2024, urban environments closely related to daily life, such as urban communities, streets, and rooftop gardens, have become hotspots in CES research. 4) In the early stages of research, CES assessment primarily relied on monetary economic methods and manual evaluation methods. With the growing demand for big data analysis, ecological analysis models have been widely applied, and artificial intelligence technologies such as machine learning and deep learning have emerged as the latest assessment methods. When addressing the massive demand for data analysis, emerging machine learning and deep learning models facilitate the processing of large datasets and the extraction of in-depth information, significantly enhancing the efficiency and accuracy of CES assessment research. Among these, CES assessment methods based on natural language processing (NLP) and computer vision (CV) recognition technologies are particularly representative and have become a hot research focus both domestically and internationally in recent years. Specifically, classic deep learning models such as ResNet, EfficientNet, YOLO, and BERT, as well as emerging large language models like GPT and Gemini, are among the most frequently used assessment tools. 5) This study established a CES assessment framework based on big data, forming an expandable and transferable standardized assessment workflow through a cascading mechanism of “assessment object−CES value type−big data type−assessment method”, providing an innovative paradigm for ecosystem service research of different scales, scopes, and types.

Conclusion In summary, early CES research focused on economic value calculation and environmental quality assessment. With the increasing demand for high-quality human settlements, the research focus has gradually shifted to the socio-cultural dimension, emphasizing cultural benefits such as health benefits, identity recognition, and spiritual value, driving CES research into a new phase of human well-being and perception assessment. Future research should strengthen the application of multi-source big data integration and interdisciplinary methods, with a focus on constructing standardized CES assessment frameworks to enhance their theoretical explanatory power.