Web servicesNational Library of Medicine (NIH / NCBI)
PubMed
Decision matrix
PubMed app decision matrix
Overview
PubMed is the free biomedical literature database from the US National Library of Medicine, with a machine-learning Best Match ranking and export straight into a citation manager. Pick it for any search of the published literature. It is not an AI summarizer, so pair it with a chat assistant only after you have read the actual abstracts.
Other good fits
Exporting a search's results into a reference manager like EndNote or Zotero as a .nbib file.
Setting up a free My NCBI account to get email alerts when new papers match a saved search.
+ Pros
Read pros
PubMed is the free literature database the National Library of Medicine runs through NCBI, covering biomedical and life-science journals. Its default "Best Match" sort re-ranks results with a machine-learning model on top of a term-frequency algorithm, weighing citation frequency, where your terms appear in the article, and how recent it is. No account is needed to search or export; a free My NCBI account adds saved searches, email alerts and display preferences.
Pick it when
You need to find what has already been published on a topic before starting a protocol or a grant.
You want citations exported straight into EndNote, Zotero or another reference manager.
You want an automatic email whenever a new paper matches a search you care about, via a free My NCBI account.
− Cons
Read cons
PubMed indexes abstracts and metadata; it does not host every paper's full text, so you will often be redirected to a publisher's paywall. Best Match is a relevance aid, not a citation-count or impact-factor ranking, and its exact weighting is not something you control search by search, so two people searching the same terms on different days can see slightly different top results as the model and index update. It has no built-in summarisation: reading and synthesising what the papers say is still on you.
Pick something else when
You want an AI assistant to draft a summary of what several papers found: read the abstracts in PubMed first, then summarise with ChatGPT or Claude, checking every claim against the source.
You need full-text search across a specific journal or publisher's archive: use that publisher's own search, which PubMed will often link to anyway.
You need citation-network or impact-metric analysis rather than a reading list: that is a different kind of tool from a literature search engine.
Avoid if
You want an AI tool to summarise or synthesise findings across papers; PubMed ranks and finds, it does not summarise.
You need a systematic-review-grade, fully reproducible search; document your exact search string rather than relying on Best Match's default order.
Workflow
Build a search with specific terms rather than a single word → sort by Best Match for a quick read or by date for the newest papers → open promising abstracts and check what they actually say → export the citations you plan to keep as a .nbib file or CSV → save the search in My NCBI and turn on an email alert if you'll need to repeat it.
Cost
PubMed is completely free, with no paid tier, and always has been; it is funded by the US government through the NIH. There is nothing to install: it runs as a website, with an optional free My NCBI account for saved searches and alerts. The ongoing effort is search hygiene: refining terms so Best Match has something specific to rank, and periodically pruning saved alerts that have stopped being useful.
Restrictions
No account is required for core search and export, but My NCBI's saved searches, alerts and display customisation only work once you register a free account. Some linked full-text content requires an institutional or personal subscription outside PubMed itself.
Search PubMed for your topic using two or three specific terms, not one broad word.
Sort by Best Match, open the top 10 results, and note which ones are actually relevant.
Export those citations as a .nbib file and import them into your reference manager.
Create a free My NCBI account, save the search, and turn on a weekly email alert.
A week later, check whether the alert caught a paper you would have wanted anyway.
Done when the exported citations open cleanly in your reference manager and the alert reliably surfaces new, relevant papers.
Links & evidence
Documentation-based · not hands-on · Sep 23, 2026
Rests on PubMed's own User Guide and homepage, read on 2026-09-23, for the Best Match algorithm description, export formats and account requirements. No search was run in PubMed for this review; the first project is a suggested exercise.
PubMed User Guide · Sep 23, 2026 — Best Match re-ranks term-frequency results with a machine-learning algorithm; export formats include .nbib, XML, CSV, plain text; My NCBI is optional.
PubMed homepage · Sep 23, 2026 — Free biomedical literature search run by the National Library of Medicine / NCBI, no login required for basic search.
NLM: How does the PubMed Best Match feature work? · Sep 23, 2026 — Confirms Best Match is machine-learning based with 150+ ranking signals, and states directly that "Best Match is not designed for comprehensive or systematic searches."
Further review notes & alternatives
Alternatives and trade-offs
ChatGPT or Claude can help you turn a reading list you found in PubMed into a summary or comparison table, but neither is a literature index and both can invent citations if you ask them to "find papers" directly.
A publisher's own search portal can be faster for full-text search within a single journal family, at the cost of not covering the wider literature the way PubMed does.
A citation manager like Zotero or EndNote is the natural next step after PubMed for organising and formatting the references you've exported.
PubMed is the free biomedical literature database run by the US National Library of Medicine. Its Best Match relevance sort uses a machine-learning model on top of term-frequency ranking, and results export to citation managers, plain text, XML or CSV without needing an account; a free My NCBI account adds saved searches and email alerts.
National Library of Medicine (NIH / NCBI) · free · cloud