‘Two Sides of the Same Coin’: A Multi‐Stakeholder Perspective on Discoverability and Promotionality in a Comparison of Scholar‐Authored vs. AI ‐Generated Keywords From the Chinese Social Sciences

For scholarly publishers, Generative AI presents efficiency opportunities but critical metadata quality challenges. To determine if AI can replace authors, we conducted a multi‐dimensional comparative analysis of 1500 paired scholar‐authored and Large Language Model (LLM)‐generated keyword sets from high‐impact Chinese social science articles, assessing grammatical, semantic and discourse features. Supported by rigorous statistical testing, we found the two approaches represent ‘two sides of the same coin’. The AI employed a bottom‐up ‘intelligent summarization’ logic, creating comprehensive sets optimised for discoverability. Conversely, human scholars used a top‐down ‘artisanal distillation’ logic to serve promotionality by crafting conceptually novel terms. Our findings demonstrate that full automation is suboptimal, sacrificing the crucial promotional function that brands research and signals specific contributions. We conclude the ideal path is a synergistic human‐AI workflow. This critical curation model enhances metadata by creating keywords that are both comprehensive and conceptually potent.