Finding Scholarly Sources with AI

Searching for scholarship begins with a question, not with a list of references a chatbot happens to produce. PubMed’s user guide advises identifying key concepts and offers field tags, Boolean operators and other search tools; it also distinguishes a citation record from full text. Crossref provides DOI metadata that can help verify a reference. AI can help generate synonyms, broader and narrower terms, or a draft search string, but it may invent a plausible title or combine metadata from two real papers. Choose the index that fits the subject and assignment. Search with a few core concepts and inspect the results. Record why a candidate is relevant: its research question, population or materials, method and publication context. Open the actual record or full article when available. Verify the author, title, journal, year and DOI; then read enough of the paper to ensure it supports the specific claim. A correct DOI only proves a record exists, not that its conclusion matches the model’s summary. Use AI to compare search strategies rather than to replace evaluation. Ask which concepts may be missing, what alternative terminology authors use and what kind of study would answer the question. Check those suggestions in the database. Follow references and citing papers when appropriate, but beware of treating a review, preprint, editorial and original experiment as interchangeable. The library or instructor may require particular source types. Keep a reproducible note with the database, date, query and reasons for inclusion or exclusion. Do not rely on a chatbot’s bibliography alone. Where a source is paywalled, use legitimate library access or the abstract with an explicit limitation; do not imply you read the full results. The useful role for AI is search preparation and navigation, while the scholar remains responsible for selecting and representing evidence accurately.