PubMed Article Search Scraper API

Search biomedical literature through the official NCBI PubMed E-utilities. Use native PubMed query syntax or structured title, author, journal, publication-date, article-type, and language filters. Results include source-linked bibliography records; optional abstract fetching uses NCBI EFetch.

Fast first run

{"query":"artificial intelligence[Title]","sort":"newest","includeAbstract":false,"maxResults":5}

The deterministic default returns up to five recent records and costs at most $0.0101 ($0.0001 start + $0.002 per dataset item).

Output

Found rows include pmid, title, stable authors, journal/full journal name, publication and e-publication dates, DOI and PMC identifiers, stable identifier/publication-type/language arrays, exact pubmedUrl, query/sort/result-position metadata, source retrieval time, bounded-query disclosure, and mandatory evidence limits. abstract is populated only when includeAbstract=true and source abstract text is available.

NO_MATCH, INVALID_INPUT, and SOURCE_UNAVAILABLE produce explicit evidence rows rather than silent empty datasets, unless a buyer charge cap permits zero items.

Examples

Structured review search:

{
  "title": "machine learning",
  "journal": "Nature Medicine",
  "publicationDateFrom": "2025/01/01",
  "publicationDateTo": "2026/12/31",
  "articleType": "Review",
  "language": "English",
  "sort": "newest",
  "includeAbstract": true,
  "maxResults": 25
}

Native syntax:

{"query":"(diabetes[Title/Abstract]) AND randomized controlled trial[Publication Type]","sort":"relevance","maxResults":10}

maxResults is a total 1–500 item cap, not a per-page cap. Apify's Maximum total charge further reduces the cap before source enrichment.

Evidence boundaries

• PubMed is an indexed bibliographic database; inclusion is not NCBI endorsement.
• A record does not prove peer review, article quality, clinical validity, applicability, or consensus.
• Results are not clinical guidance or medical advice.
• A bounded query is not complete literature coverage. Search syntax, indexing status, filters, and source updates affect recall.
• Publication dates, identifiers, abstracts, and other metadata can be corrected or updated.
• Abstract availability varies and an empty abstract does not establish that the article has no abstract elsewhere.

Always review the linked PubMed record and underlying publication before relying on a result.

Reliability and NCBI usage

The actor identifies itself with a descriptive tool/email User-Agent, paces no-key requests below NCBI's general three-requests-per-second guidance, batches PMIDs (up to 100/request), and bounds pagination. It retries transport errors, body-read failures, HTTP 408/429/5xx responses (including Retry-After), and malformed JSON/XML responses. It never substitutes unofficial mirrors when NCBI is unavailable.

Pricing

Component	Price
Actor start	$0.0001
Dataset item	$0.002

Five items cost at most $0.0101; 25 cost at most $0.0501.