A Comprehensive Guide to Scientific Search Sources: A Four-Category Methodological Framework

AbstractSelecting an appropriate search source is arguably the most consequential methodological decision in any evidence synthesis effort, including systematic reviews, meta-analyses, qualitative meta-syntheses, scoping reviews, and bibliometric studies. Despite this, both graduate students and experienced researchers frequently restrict their searches to one or two databases, introducing substantial publication bias, database bias, and coverage loss.This article proposes a four-category framework for organizing and comparing scientificsearch sources: (1) Scientific Publishers (such as ScienceDirect, SpringerLink, IEEE Xplore), (2) Citation Databases (such as Scopus, Web of Science, Dimensions, PubMed, Embase), (3) Academic Search Engines (such as Google Scholar, Semantic Scholar, OpenAlex), and (4) AI-powered Research Tools (such as Elicit, Consensus, SciSpace, ResearchRabbit, Scite.ai). For each category, the most influential platforms are examined in depth, presenting comparison tables, strengths, weaknesses, and specific applications. Furthermore, the methodological reasons why general large language models (LLMs) such as ChatGPT, Gemini, and Claude are unsuitable for systematic searching are elucidated. Designed as a pedagogical reference for literature review and bibliometric workshops, this article aims to provide practical guidance on selecting a robust multi-database search strategy compliant with PRISMA-S guidelines.