Google Scholar Search Scraper - Citations & PDFs, No Login

Search Google Scholar with your own queries and get one Dataset row per article: title, link, authors, venue, year, snippet, cited-by count and a direct PDF or HTML link when Scholar shows one. Add a year range, skip [CITATION] entries or sort by date. No Google login, account or API key is needed. The prefilled example returns 10 articles in under 15 seconds and costs $0.010 on the FREE tier.

What it does · Quick start · Use cases · Input · Output · Pricing · Limits · API and AI agents · Troubleshooting · FAQ · Related Actors · Support

What it does

The Actor opens the public Google Scholar results pages for each query, page by page (10 articles per page), and reads every result. It returns up to 100 articles per query in Scholar's order, with positions that continue across pages. Your year range, citation and sort settings are passed to Scholar as filters. Duplicate articles inside one query are removed by Scholar's own result id (resultId).

Good fit

• Literature review: the top articles on a topic, with authors, venue, year and citation counts in one spreadsheet.
• Finding open full texts: pdfUrl gives the direct [PDF] or [HTML] link from the right column of Scholar.
• Comparing topics or periods: several queries in one run, each with the same year range.
• Feeding papers to an AI agent or a RAG pipeline: titles, snippets and links as clean JSON.

Not a fit / limits

• More than 100 articles per query. Scholar cuts deeper pages for automated traffic; split a big topic into narrower queries or year ranges.
• Author profile pages, "Cited by" lists, BibTeX/citation export or full abstracts. The Actor returns the snippet Scholar shows, not the abstract.
• Downloading PDFs. The Actor returns links; it does not fetch or store files.
• Pages behind a CAPTCHA. The Actor never solves CAPTCHAs. When Scholar blocks every proxy IP, the query gets an error row and is not charged.

Quick start

1. Open the Input tab. The form is prefilled with the query web scraping and Max results per query = 10.
2. Click Start.
3. Open the Results view in the Dataset. Check title, authors, year, citedBy and pdfUrl.
4. Open the RUN_SUMMARY record in the Key-value store to see the run status and counts.

{
  "queries": ["web scraping"],
  "maxResults": 10
}

Expected result: 10 rows with status complete, positions 1–10, from the first Scholar results page for "web scraping".

Use cases

Buyer job	Input path	Output entity	Key fields	Next step
Top papers on a topic	queries + maxResults	Article	title, authors, year, citedBy	Export CSV or Excel
Papers from a fixed period	yearFrom + yearTo	Article	year, venue, url	Screening spreadsheet
Open full texts	includeCitations: false	Article	pdfUrl, fullTextType, source	Download with your own tool
Newest papers	sortByDate: true	Article	addedAgo, title, url	Read via API
Papers by one author or journal	queries with author: or source:	Article	authorProfiles, venue, citedBy	Join on resultId

Input

Field	Type	Required	Default / prefill	Allowed values	Effect
queries	array of strings	yes	prefill ["web scraping"]	1–200 queries	One Scholar search for each line. Scholar operators work, e.g. author:"J Smith", source:Nature, "exact phrase". Empty lines and duplicates are removed.
maxResults	integer	no	default 10, prefill 10	1–100	Articles per query. The Actor stops earlier when Scholar has no more results.
yearFrom	integer	no	none	1000–2100	Only articles from this year or later.
yearTo	integer	no	none	1000–2100	Only articles from this year or earlier. Must not be lower than yearFrom.
includeCitations	boolean	no	true	true, false	false drops [CITATION] entries: works Scholar knows only from reference lists, usually without a link.
sortByDate	boolean	no	false	true, false	true sorts newest first, like Scholar's "Sort by date". Scholar then shows only articles added in the last year.
proxyConfiguration	object	no	Apify Proxy (datacenter)	Apify proxy groups or your own proxy URLs	See below.

A run with an empty queries list, maxResults above 100 or yearFrom > yearTo is rejected before any request to Scholar.

Proxy. With the default setting the Actor starts on Apify datacenter IPs. When Scholar refuses or shortens a page there, the Actor retries it through Apify residential proxy (US) and keeps that IP for the next page of the same query. If you choose a proxy group or your own proxy URLs, only your proxy is used. Apify Proxy traffic is included in the price; you pay no extra for it.

Input example

{
  "queries": ["CRISPR gene editing", "machine learning interpretability"],
  "yearFrom": 2020,
  "yearTo": 2023,
  "includeCitations": false,
  "maxResults": 25
}

Expected behavior: two Scholar searches limited to 2020–2023, without [CITATION] entries, three pages each. Expected output: up to 50 article rows with year between 2020 and 2023 where Scholar shows a year. Cost on FREE: $0.001 + 2 × $0.002 + 50 × $0.0007 = $0.040.

Output

Each Dataset row is one article (status: "complete"), or one status row for a query that found nothing (not_found, charged only as a query) or failed (error, free). The Results view shows the main columns; the JSON, CSV and Excel exports contain every field.

Field	Meaning	Type	Can be empty?
query	Your query	string	no
position	Place among the articles returned for this query; page 2 normally starts at 11 (repeats from earlier pages are skipped)	integer	null on status rows
page	Scholar results page (1 = first)	integer
resultId	Scholar's own id of the result; stable across runs	string	null on status rows
type	article, pdf, html, book or citation, from Scholar's label	string	null on status rows
title	Article title	string	null on status rows
url	Link of the title (publisher page or repository)	string	null for [CITATION] entries without a link
authors	Author names as Scholar prints them (initials and surname)	array	can be []
authorsTruncated	true when Scholar cut the author list with "…"	boolean
authorProfiles	Authors with a Scholar profile: name, profileId, profileUrl	array	can be []
venue	Journal, conference or book; Scholar may shorten it with "…"	string	null when Scholar shows none
year	Publication year	integer	null when Scholar shows no year
source	Domain or publisher at the end of the result line, e.g. nature.com	string	yes
snippet	Text excerpt under the title	string	yes
addedAgo	With sortByDate only: how long ago Scholar added the article	string	null otherwise
citedBy	Number from "Cited by N"	integer	see below
citedByUrl	Scholar page with the citing articles	string	yes
versions	Number from "All N versions"	integer	see below
versionsUrl, relatedUrl	Scholar "All versions" and "Related articles" pages	string	yes
pdfUrl	Direct full-text link from the right column	string	null when Scholar shows none
fullTextType	Label Scholar shows for that link: usually PDF or HTML, sometimes another type such as PS	string	null without pdfUrl
searchUrl	Scholar results page this row came from	string	no
status	complete, not_found or error	string	no
statusReason, errorCode	Why a row is not_found or error; code such as SOURCE_BLOCKED	string	null on article rows
fetchedAt	When the page was fetched (ISO 8601, UTC)	string	no

null is not 0. citedBy: 0 means Scholar shows the link bar of the result but no "Cited by" link. citedBy: null means the link bar is missing, so the count is unknown. year: null means Scholar shows no year; it is never written as 0.

Output example

One row from the prefilled run, shortened (relatedUrl, versionsUrl and citedByUrl removed):

{
  "query": "web scraping",
  "position": 1,
  "page": 1,
  "resultId": "vzUZDE4wNWMJ",
  "type": "pdf",
  "title": "Web scraping or web crawling: State of art, techniques, approaches and application.",
  "url": "http://i-csrs.org/Volumes/ijasca/2021.3.11.pdf",
  "authors": ["MA Khder"],
  "authorsTruncated": false,
  "authorProfiles": [
    { "name": "MA Khder", "profileId": "BUTYh5gAAAAJ", "profileUrl": "https://scholar.google.com/citations?user=BUTYh5gAAAAJ&hl=en" }
  ],
  "venue": "International Journal of Advances in Soft Computing & …",
  "year": 2021,
  "source": "i-csrs.org",
  "snippet": "… are required to implement web scrapping properly such as … into what web scraping is, how it works, web scraping stages, …",
  "addedAgo": null,
  "citedBy": 679,
  "versions": 4,
  "pdfUrl": "http://i-csrs.org/Volumes/ijasca/2021.3.11.pdf",
  "fullTextType": "PDF",
  "searchUrl": "https://scholar.google.com/scholar?q=web+scraping&hl=en&as_sdt=0%2C5",
  "status": "complete",
  "statusReason": null,
  "errorCode": null,
  "fetchedAt": "2026-10-17T12:32:46.110Z"
}

Scholar results change over time, so the same query can return a different order or citation count tomorrow.

Status rows and run summary

Situation	What you see	What to do
Articles found	complete rows	Use them
Scholar found nothing for the query and filters	One not_found row with statusReason	Check spelling, operators or the year range
A page failed after some articles	The articles, then one error row: "Stopped after N result(s)"	Run the query again or narrow it
Scholar blocked the query on every IP	One error row with errorCode SOURCE_BLOCKED	Try again later
Scholar answered no new article on 16 new IPs in a row	The run stops early; open queries get error rows with SOURCE_BLOCKED	Try again later; you paid only for articles already saved

The Key-value store record RUN_SUMMARY has the run status (COMPLETE, PARTIAL, CAPPED, NO_RESULTS, ERROR or INVALID_INPUT), the stop reason, counts (including repeatsSkipped: articles Scholar repeated from earlier pages, which are not saved again) and one line per query with its result count and Scholar's "About N results" total. A "did not match" answer is checked once more on a new IP before the query is reported as not_found.

Pricing

Pay-per-event. You pay for three things:

Event	When	FREE	BRONZE	SILVER	GOLD and above
Actor start	Once per run	$0.001	$0.001	$0.001	$0.001
Query	Once per query that Scholar answered (articles or not_found)	$0.002	$0.002	$0.002	$0.002
Result	Each complete article row	$0.0007	$0.00063	$0.00056	$0.00049

error rows are free, and a query that fails before Scholar answers is not charged. Platform usage (compute, proxy, storage) is included.

Scenario	Input	Charged events	FREE	GOLD
Prefill	1 query × 10	start + 1 query + 10 results	$0.010	$0.008
One deep query	1 query × 100	start + 1 query + 100 results	$0.073	$0.052
1,000 articles	10 queries × 100 (needs a longer timeout)	start + 10 queries + 1,000 results	$0.721	$0.511
Query with no articles	1 query, not_found	start + 1 query	$0.003	$0.003

Your maximum cost per run stops the run early when the next article or query would go over it: the run then ends SUCCEEDED with status CAPPED and reason CHARGE_LIMIT, keeping the rows that fit. Apify controls actual billing; see the Pricing tab for your tier.

Limits

• Up to 200 queries per run and 100 articles per query. In our tests 200 queries × 1 article took about 9 minutes with a 600-second timeout, so raise the timeout for big lists.
• Speed: a 10-article run usually takes under 10 seconds. In our tests 3 queries × 50 articles took about 30 seconds, and 5 queries × 100 articles took 2–3 minutes at 256 MB.
• Deep pages: Scholar often repeats or empties pages after about 70 articles of one query. In our tests 1–2 of 5 queries × 100 ended with a free error row after 72–92 articles; the other queries reached 100.
• The default run timeout is 300 seconds. The Actor stops source work 15 seconds before the timeout and keeps what it has: the run ends SUCCEEDED with status CAPPED and reason DEADLINE. For many queries, raise the timeout in the run options.
• Default memory is 256 MB; more memory does not add results.
• Results are in English Scholar (hl=en) with Scholar's default search scope (articles, no patents).

This Actor is independent of Google. It reads public Google Scholar results pages, the same pages you see in a browser without login. Articles belong to their authors and publishers; check their terms before you reuse texts or PDFs.

API and AI agents

Run the Actor with the same JSON input through the Apify API. Keep your Apify token private. Read the default Dataset items and the RUN_SUMMARY record from the finished run. Apify exports JSON, CSV and Excel and can run the Actor on your own schedule.

Use with AI agents (MCP)

• MCP server with only this Actor: https://mcp.apify.com?tools=cliqtomedia/google-scholar-scraper (authenticate with Apify).
• Smallest useful input: {"queries": ["web scraping"], "maxResults": 10}.
• Key output fields: title (paper title), authors (names), year (publication year or null), citedBy (citation count or null), pdfUrl (direct full text or null).
• A 10-article call usually finishes in under 15 seconds and costs $0.010 on FREE. Limits: 100 articles per query, 200 queries per run.

Troubleshooting

A query returns a not_found row, but Scholar in my browser shows articles

Likely reason: a different spelling, operator or year range, or a filter such as includeCitations: false. Check the searchUrl of the row in a browser. Action: fix the query or remove a filter.

The last row of a query is an error row

Likely reason: Scholar refused or cut a deeper page on every proxy IP. The articles before it are kept and paid; the error row is free. Action: run the query again, or ask for fewer articles per query.

The run ends with CAPPED and DEADLINE

Likely reason: many queries for the run timeout. Action: raise the timeout in the run options, or split the queries into several runs.

FAQ

Do I need a Google account or a SerpApi key?

No. The Actor reads public Scholar pages through Apify Proxy. You only need an Apify account.

Why is citedBy or year empty for some rows?

Scholar does not show them for every result. null means "not shown", not zero. [CITATION] entries often have no year and no link.

Can I get every article on a topic?

No. Scholar shows at most about 1,000 results for one search, and it cuts deep pages for automated traffic; this Actor returns up to 100 per query. Use narrower queries or year ranges to cover more.

Are the results the same as in my browser?

Not always. Scholar answers different IPs and sessions differently: when we ran the same query several times, the order and some of the articles changed between runs. position is Scholar's order at the time of the run, so compare by title or url, not by position.

Related Actors

• DOI Metadata & Bibliography Enrichment — turn the papers you found into full bibliographic records by DOI.
• Google Patent Details Scraper — the same research step for patents.

Support

Use the Actor's Issues tab on Apify. Include the run ID, the expected result and a non-sensitive input example. Never share tokens or cookies.