literature-review

A systematic literature-review workflow for academic, biomedical, technical, and scientific topics, spanning search planning, source screening, synthesis, citation verification, and reproducible evidence logging. It suits building narrative, scoping, systematic, or meta-analytic reviews, mapping the state of the art, finding gaps or contradictions, and preparing citation-backed background sections. When rigor is unspecified, the default is a scoping review for exploratory work and a systematic review for publication or clinical claims.

The process is staged. First, the prompt is converted into a searchable question, using PICO (Population, Intervention, Comparator, Outcome) for clinical or biomedical work, or a system, method, baseline, and metric framing for technical work. A search protocol is written before collecting sources, fixing databases, date range, languages, publication types, inclusion and exclusion criteria, and exact search strings. The recommended minimum database set is PubMed for biomedical literature, arXiv for CS, math, physics, and preprints, and Semantic Scholar or Crossref for broad discovery, plus domain-specific registries or standards when relevant.

A reproducible search log records each query and result count. Deduplication proceeds by DOI, then PMID or arXiv ID, then exact title, then normalized title with first author and year. Screening runs in title, abstract, and full-text stages with recorded exclusion reasons, and data is captured in a structured extraction table. Synthesis groups evidence by theme and separates claims into high, medium, and low confidence rather than summarizing papers one at a time. A citation-verification step checks identifiers, authors, and years, marks preprints, and forbids citing a paper for a claim it does not make. An output template and a list of pitfalls, such as not treating search snippets as evidence, close the workflow.