A defensible literature search is more than a question entered into one search box. It requires explicit concepts, sources with suitable coverage, recorded queries, and citation-based checks for omissions. AI can expand terminology and rank candidates; it cannot establish that a search is complete.
When this workflow is useful
- Entering an unfamiliar topic and mapping the field
- Finding evidence for a proposal, review, or experiment
- Tracing the history and later use of a key paper
A practical sequence
- 01
Define the question and scope
State the object, phenomenon, method, period, and exclusions. Exploratory searches can remain broad; systematic searches require criteria specified in advance.
- 02
Build a terminology table
List synonyms, historical names, abbreviations, spelling variants, and domain vocabulary. AI can suggest candidates, but researchers should confirm them against known papers and disciplinary indexes.
- 03
Use complementary routes
Combine subject, author, backward-citation, and forward-citation searches in sources appropriate to the field. No single scholarly index has universal coverage.
- 04
Screen, iterate, and record
Inspect relevant abstracts, keywords, and references. Revise queries as vocabulary improves, preserving the platform, date, query, screening reason, and stable identifiers.
Quality-control questions
- Do sources cover the relevant discipline, document types, and period?
- Could another researcher reconstruct the search?
- Were consequential statements checked in original papers?
- Was coverage challenged through citation tracking or another source?