Optimizing scholarly literature search: a comparative study of relevance and evidence quality across AI-powered, semantic, and traditional search engines

Purpose Artificial intelligence (AI)-powered search engines are increasingly used for evidence retrieval, but their effectiveness in complex, interdisciplinary fields like tissue engineering is underexplored. This study aims to evaluate the performance of Consensus, Semantic Scholar and Google Scholar in retrieving relevant literature. Design/methodology/approach Custom search strategies were applied: PICO queries for Google Scholar, natural language input for Semantic Scholar and targeted prompting for Consensus. The top 50 results per platform were assessed for precision, false drop rate, study design, temporal distribution and citation metrics. Relevance was judged by experts using fuzzy logic, and evidence quality was evaluated against medical hierarchies. Findings Consensus achieved the highest performance with 86.0% precision and no false drops, outperforming Semantic Scholar (75.5% precision, 10.0% false drop rate) and Google Scholar (72.0% precision, 10.0% false drop rate). It also retrieved more highly cited articles (average 341 citations) and showed a moderate positive correlation (rs = 0.65) between relevance and citation frequency. Semantic Scholar yielded slightly newer publications, but no significant differences were observed in study type or temporal distribution. Duplicate retrievals were minimal (8%). Originality/value Purpose-built AI tools like Consensus enhance precision and impact in evidence retrieval, offering strong potential for interdisciplinary biomedical research.