Conservation and restoration of cultural heritage are facing more serious problems today due to the degradation of materials, environmental risks, fragmentation of historical evidence, and high demands for data-driven decisions. Although artificial intelligence (AI) has been widely applied in heritage research, most of the existing studies have focused on individual technologies, specific heritage objects or isolated application scenarios, and thus have not comprehensively explored the general development and conservation-oriented value of this field. Bibliometric analysis of applications of AI in cultural heritage conservation and restoration from 2015 to 2025 is conducted in this paper. CiteSpace, VOSviewer and Bibliometrix were used to study changes in publication trends and co-occurrence relationships of related works over time. The above results show that the three stages of development for this field are: exploratory, stable growth and fast development. Research has been conducting damage detection, image completion, generative reconstruction, structural monitoring and preventive conservation recently, away from digital documentation and technical feasibility studies. In addition to mapping the knowledge structure of this field, the research has also connected AI applications to the three restoration-oriented paths and proposed a “restoration task – data type – evidence standard” framework for interpreting AI-assisted heritage restoration. It is proposed that the future application of artificial intelligence (AI) in heritage conservation will be achieved by linking the results of algorithms with other elements such as historical knowledge, expert assessments, and conservation measures on actual heritage objects. This paper reviews the macro-level situation of AI-assisted heritage restoration and offers references for future interdisciplinary research and conservation workflow development.
