关键词: 学术搜索平台, 生成式人工智能, 大语言模型, 生成式检索, 增强检索

Abstract: The study explores the integration trend of generative artificial intelligence (GenAI) technology and academic search, and comprehensively evaluates the performance of emerging academic search platforms in practical applications, aiming to provide a new pathway for transformation in the field of information retrieval. Four GenAI-powered academic search platforms, Scopus AI, WoS Research Assistant, SciSpace, and Elicit, are selected to construct the evaluation index system on the dimensions of content generation and retrieval generation. The study reveals their differences in terms of literature coverage, generation accuracy and user experience, and further tests the balance between citation recall and precision rate of GenAI. It is found that the four GenAI academic search platforms are smooth and informative in response, but the coverage and precision of the generated content supported by citations need to be further improved. In comparison, Scopus AI performs relatively better in literature coverage, while WoS Research Assistant stands out in terms of precision rate. SciSpace achieves a more balanced performance in cross-disciplinary adaptability, whereas Elicit shows an overall balanced but slightly weaker performance. Although the fluency of these platforms is high, the problem of “information hallucination” still exists, and the accuracy and reliability of the generated content need to be further improved in the future.