Semantic Scholar Scraper — Papers, Authors & Citations

Search Semantic Scholar and retrieve structured scholarly metadata for literature reviews, research discovery, citation mapping, bibliometrics, and AI-agent research workflows. The actor uses the public Semantic Scholar Academic Graph API rather than browser HTML, so results are compact, typed, and easy to page through.

What it returns

Each dataset row has one stable shape and a recordType of paper, author, citation, or reference.

Field	Meaning
paperId, authorId	Semantic Scholar identifiers
title, abstract, tldr	Paper title/abstract/short summary, or author name
authors	Compact author ID/name objects
citationCount, referenceCount, influentialCitationCount	Graph metrics when supplied
externalIds, url, openAccessPdf	DOI, arXiv, canonical URL, and OA PDF metadata
venue, year, publicationDate	Publication metadata
hIndex, paperCount	Author metrics when supplied
relation	cites or references for graph traversal rows

When to use

Use searchPapers to discover papers by topic, getPaper for one known work, searchAuthors or getAuthor for researcher metadata, and getCitations/getReferences to traverse a paper’s graph. This actor does not download PDFs, extract full text, or bypass private publisher access; use a full-text or document actor for those jobs.

Input examples

Paper search:

{ "operation": "searchPapers", "query": "retrieval augmented generation", "year": "2020-2024", "limit": 25, "openAccessPdfOnly": true }

Paper details:

{ "operation": "getPaper", "paperId": "10.1038/s41586-020-2649-2", "fields": ["paperId", "title", "authors", "abstract", "citationCount", "openAccessPdf"] }

Citation graph:

{ "operation": "getCitations", "paperId": "649def34f8be52c8b66281af98ae884c09aef38b", "limit": 50, "offset": 0 }

For AI agents, the operation description is the routing boundary: use paper search for discovery, one-paper lookup for known identifiers, and graph operations for citation relationships. Keep limit tight and paginate with offset to control context and cost.

Pricing

Event	Price
Returned record	$0.001 per dataset row

The worst-case charge is limit × $0.001 for list operations. A one-paper lookup costs one record event. Semantic Scholar API traffic is not a browser scrape; no residential proxy is required.

Reliability and limits

The actor retries transient API errors and HTTP 429 responses with exponential backoff. A valid query with no matches is a successful run with zero rows. Invalid input, missing identifiers, API rate limiting, and upstream errors are reported in the OUTPUT key-value record with actionable warnings instead of being presented as a misleading empty search.

Semantic Scholar documents introductory API-key rate limits of approximately one request per second. Set the SEMANTIC_SCHOLAR_API_KEY environment variable in Apify for higher limits when available. The API can omit abstracts, TLDRs, PDFs, dates, and external IDs for individual records; those fields are returned as null or empty arrays consistently.

Legal and responsible use

Use the Semantic Scholar API according to its current terms, attribution requirements, rate limits, and applicable law. Respect publisher copyright and do not represent API metadata as a license to redistribute full papers.