Google Scholar Scraper (Papers, Citations, Free PDFs)

What is Google Scholar Scraper?

Google Scholar Scraper collects Google Scholar search results for any topic. For each paper it returns the title, link, authors, venue, year, a snippet, the cited-by count, the number of versions and a free PDF or HTML link when Scholar shows one. Use it for literature reviews, citation tracking, lists of papers per topic and research monitoring. It reads public Scholar result pages. No Google login is used, and no cookies are sent.

Each query walks Scholar's result pages (10 papers per page) until your result limit is reached or Scholar has no more results. One dataset row is one paper, and you pay per row.

What data does Google Scholar Scraper extract?

Field	Type	Meaning
resultId	string	Scholar's own id for the paper (stable across runs)
position	number	Rank in the query's results (1 = first)
page	number	Scholar result page the row came from (10 per page)
query	string	The search query
title	string	Paper title, without the [PDF] or [BOOK] label
documentType	string or null	Scholar label such as BOOK, HTML, CITATION, when shown
url	string or null	Link to the publisher or source page
authors	string or null	Author list as Scholar prints it (often abbreviated, e.g. "R Van Noorden")
venue	string or null	Journal, conference or publisher line
year	number or null	Publication year
source	string or null	Domain or publisher shown by Scholar
snippet	string or null	Text excerpt with the matched terms
citedBy	number	Cited-by count (0 when Scholar shows none)
citedByUrl	string or null	Scholar page listing the papers that cite this one
versionsCount	number	Number of Scholar versions of the paper
versionsUrl	string or null	Scholar page listing all versions
freeAccessType	string or null	PDF or HTML, when Scholar shows a free copy
freeAccessUrl	string or null	The free copy link
scrapedAt	string	ISO timestamp of the scrape

How to use Google Scholar Scraper

1. Open the Actor and enter one or more queries in "Search queries", one per line, e.g. solid-state battery.
2. Set "Max results per query" (default 20, up to 100).
3. Optional: set "Published from" and "Published to" years, or turn on "Sort by date".
4. Leave "Connection mode" on Auto. It tries a direct request first and switches to a residential proxy only after Scholar blocks it.
5. Start the run, then download the dataset as JSON, CSV or Excel, or read it through the API.

Example input

{
    "queries": ["solid-state battery"],
    "maxResultsPerQuery": 20
}

Example output

Real row from a run on 23 Oct 2026 (query federated learning, residential proxy):

{
    "resultId": "FE0GVCg5twYJ",
    "position": 1,
    "page": 1,
    "query": "federated learning",
    "title": "A review of applications in federated learning",
    "documentType": null,
    "url": "https://www.sciencedirect.com/science/article/pii/S0360835220305532",
    "authors": "L Li, Y Fan, M Tse , KY Lin",
    "venue": "Computers & Industrial Engineering",
    "year": 2020,
    "source": "Elsevier",
    "snippet": "… This TTF consists of FL API and Federated Core (FC) API. In detail, FL API offers a set of …",
    "citedBy": 2339,
    "citedByUrl": "https://scholar.google.com/scholar?cites=483918330336791828&as_sdt=2005&sciodt=0,5&hl=en",
    "versionsCount": 6,
    "versionsUrl": "https://scholar.google.com/scholar?cluster=483918330336791828&hl=en&as_sdt=0,5",
    "freeAccessType": "PDF",
    "freeAccessUrl": "https://orca.cardiff.ac.uk/id/eprint/134968/1/CAIE_A%20review%20of%20applications%20in%20federated%20learning_deposit.pdf",
    "scrapedAt": "2026-10-23T10:02:44.909Z"
}

Input parameters

Parameter	Default	Description
queries	solid-state battery	Topics or exact phrases, one per line, up to 20 per run
maxResultsPerQuery	20	Papers per query, 1 to 100
yearFrom	empty	Earliest publication year (four digits)
yearTo	empty	Latest publication year (four digits)
sortByDate	false	Newest papers first instead of Scholar relevance
proxyMode	auto	auto, direct (never proxy) or residential (always proxy)

Pricing

Pay per result: $2.00 per 1,000 papers on the free plan ($0.002 per paper). Volume tiers lower the price on paid Apify plans (down to $0.0008 per paper on Diamond).

Platform cost measured on 23 Oct 2026 at 256 MB: one run for 30 papers used $0.001 of platform usage, and one run for 10 papers used $0.002 (the CLI prints usage to three decimals, so these are rounded figures). Retries that wait on blocked requests add run time, and so cost.

Google Scholar Scraper vs. alternatives

This Actor returns Google Scholar's own ranking, cited-by counts and version lists, so it fits when those are the numbers you need. For a complete bibliographic index with a documented API, use a scholarly index service instead; this Actor does not replace one.

Using Google Scholar Scraper with AI agents / MCP

The Actor is available through the Apify MCP server and the Apify API. An agent sends queries and maxResultsPerQuery and reads the dataset rows. Keep maxResultsPerQuery small per query when an agent runs many queries, because the total charge is capped by the caller's maxTotalChargeUsd.

FAQ

Is this legal? The Actor reads public Scholar result pages, the same pages a person sees in a browser. It does not log in and does not bypass a login. Google's terms restrict automated access, so check them for your use case. You are responsible for how you use the data.

Why did a query return fewer results than I asked for? Scholar stops after a few pages for some queries, and some requests are soft-blocked. The Actor retries on fresh proxy sessions with a pause between attempts. Measured on 23 Oct 2026, transformer attention (100 requested, from 2020) returned 8 papers before a block, solid-state battery (100 requested, from 2020) returned 50, and federated learning (30 requested) returned all 30. Rerun the query later, or split the topic by publication year, to reach more papers.

Why did a query return 0 results? Scholar sometimes answers a valid query with the "did not match any articles" page, which looks the same as a real empty result. In a test on 23 Oct 2026, the unquoted graph neural network returned 0 papers from both the direct connection and the residential proxy, while the quoted phrase returned results from the same network. Try an exact phrase in quotes, or rerun the query later.

Why does the direct connection fail? From Apify datacenter IPs, Scholar answered HTTP 403 to direct requests in both direct-only test runs. Auto mode therefore switches to the residential proxy after one blocked request.

Can I call it from n8n, Make or Zapier? Those tools can start any Apify Actor through their Apify integrations. This was not tested for this Actor.

What about errors and limits? A run fails only when no paper is collected at all. If some queries are blocked, the run succeeds and logs which queries were blocked. Each run is capped by the caller's maximum charge, and the Actor stops adding rows when the cap is reached.

Related Actors

Other Actors from the same publisher are listed on its Apify profile.

Support

Open an issue on the Actor's Issues tab on Apify. Include the run ID, the input you used and the log lines that mention "Blocked".

Limitations

• Scholar's public pages return at most a few pages for some queries before blocking.
• Scholar's authors line is abbreviated ("R Van Noorden"), and it is not a DOI or ORCID. Use the url to find the full record.
• freeAccessUrl is the link Scholar shows, which can be a repository copy or a preprint. It is not checked for licence terms.
• Results are Scholar's ranking at scrape time. They change often, and two runs can differ.