Google Scholar Scraper

🎓 Google Scholar Scraper

Export Google Scholar search results, "cited by" lists and author profiles as clean, structured data. Every paper row has the title, link, authors (with Scholar profile ids), venue, year, abstract snippet, citation count, versions and the free PDF link when Scholar has one. Author rows add citations, h-index, i10-index, citations per year, co-authors and articles. No login, no API key.

✨ Why use Google Scholar Scraper?

• 🔎 Three modes in one run: keyword searches, "cited by" lists and author profiles. Mix them freely.
• 🧹 Clean fields: authors as a list, year and venue split out, citation counts as numbers, the snippet as plain text without HTML. Ready for spreadsheets, databases and AI agents.
• 📄 Free full text: the [PDF] / [HTML] link Scholar shows next to a result, with its host.
• 🧑‍🔬 Author metrics: all-time and recent citations, h-index and i10-index, a citations-per-year series, interests, verified email domain, homepage, co-authors and up to thousands of articles.
• 🔁 Handles Scholar's throttling: pages that come back short are retried on fresh IPs automatically, so you get complete pages.
• 💸 Pay per result: you pay for saved papers and profiles, never for retries, errors or empty searches.

🎯 Use cases

• 📚 Literature reviews: pull the top papers for a topic and year range into a spreadsheet in seconds.
• 🔗 Citation tracking: list every paper citing your work or a competitor's key paper.
• 📈 Bibliometrics: compare researchers by h-index, i10-index and citation growth per year.
• 🤖 AI agents and RAG: give an assistant fresh, structured scholarly results with links and PDFs to read.
• 🏛️ Grant and hiring research: profile candidates or reviewers with their affiliation, interests and publication record.
• 🧪 Trend monitoring: sort by date to catch new papers on a topic every week.

🚀 How to use

1. Type one or more Search queries (Scholar operators like author:"hinton" or source:nature work), or paste Scholar search links.
2. Optionally add Author profiles (ids like JicYPdAAAAAJ or profile links) and Cited by ids (the clusterId of any paper row, or a "Cited by" link).
3. Set Maximum papers per search and any filters (year range, sort, languages, patents, review articles).
4. Click Start, then download the dataset as JSON, CSV, Excel or HTML, or read it through the API.

📥 Input

{
  "queries": ["large language models"],
  "citedBy": ["4731455256244956465"],
  "authors": ["JicYPdAAAAAJ"],
  "maxResults": 50,
  "yearFrom": 2023
}

📤 Output

Paper rows (type: "paper"):

Author rows (type: "author"): authorId, name, affiliation, verifiedEmailDomain, homepage, interests, photoUrl, citationsAll, citationsRecent, hIndexAll, hIndexRecent, i10IndexAll, i10IndexRecent, recentSinceYear, citationsPerYear, publicAccess, coAuthors, articlesCount and articles (title, year, venue, authors, citedByCount, citesIds).

{
  "type": "paper",
  "searchType": "citedBy",
  "citedById": "4731455256244956465",
  "position": 2,
  "resultId": "2LiXtEmo2fAJ",
  "clusterId": "17355087673564838104",
  "title": "ViT: Unlocking Test-Time Training in Vision",
  "url": "https://openaccess.thecvf.com/content/CVPR2026/html/Han_ViT3_Unlocking_Test-Time_Training_in_Vision_CVPR_2026_paper.html",
  "documentType": "article",
  "authors": ["D Han", "Y Li", "T Li", "Z Cao", "Z Wang"],
  "authorsTruncated": true,
  "publication": "Proceedings of the …",
  "year": 2026,
  "source": "openaccess.thecvf.com",
  "snippet": "Abstract Test-Time Training (TTT) has recently emerged as a promising direction for efficient sequence modeling. …",
  "citedByCount": 21,
  "versionsCount": 5,
  "fullTextUrl": "https://openaccess.thecvf.com/content/CVPR2026/papers/Han_ViT3_Unlocking_Test-Time_Training_in_Vision_CVPR_2026_paper.pdf",
  "fullTextType": "pdf",
  "fullTextSource": "thecvf.com"
}

{
  "type": "author",
  "authorId": "AbCdEfGhIjKL",
  "name": "Jane Example",
  "affiliation": "Professor of Computer Science, Example University",
  "verifiedEmailDomain": "example.edu",
  "interests": ["machine learning", "computer vision"],
  "citationsAll": 15230,
  "citationsRecent": 9120,
  "hIndexAll": 48,
  "hIndexRecent": 39,
  "i10IndexAll": 96,
  "i10IndexRecent": 80,
  "recentSinceYear": 2021,
  "citationsPerYear": {"2024": 2100, "2025": 2480, "2026": 1900},
  "coAuthors": [{"name": "John Sample", "authorId": "ZyXwVuTsRqPo", "affiliation": "Example Institute"}],
  "articlesCount": 20,
  "articles": [{"title": "An example paper", "year": 2019, "publication": "Example Journal 12 (3)", "citedByCount": 1830}]
}

💰 Pricing

• $1.00 per 1,000 papers (search and cited-by rows) and $10 per 1,000 author profiles ($0.01 each, including up to the articles you ask for).
• Example: 1,000 papers ≈ $1.00 in result fees plus a few cents of platform usage (compute and datacenter traffic).
• Retries, throttled pages and failed inputs are never charged.

❓ FAQ

Is it legal to scrape Google Scholar? The actor reads public pages that anyone can open without an account. Check Google's terms and the rules that apply to your use, especially for personal data such as author profiles.

Do I need a Google account, API key or proxy? No. The default proxy setting works; it uses datacenter IPs and switches to US residential IPs for pages Google throttles.

How many results can I get? Scholar lists at most 1,000 results per search, and for signed-out visitors it throttles deep pages hard: the first 20 to 50 results of a search come back reliably, deeper pages often do not. For larger sets, run several narrower searches in one run (for example the same query with different year ranges or extra keywords); each search's first pages load reliably.

Why is a field empty? Scholar does not show everything for every result: citation-only records have no link, many results have no free PDF, and venue names are often shortened ("Proceedings of the …") by Scholar itself.

Can I get full abstracts? Scholar shows a snippet of the abstract, which is what snippet contains. For full abstracts by DOI, pair this actor with an OpenAlex scraper.

Can I schedule it or call it from code or an AI agent? Yes: run it on a schedule, through the REST API, or as a tool through MCP. Inputs are plain strings and numbers and every row is flat JSON with clear field names.

⚠️ Limits

• Scholar lists at most 1,000 results per search and throttles deep pages for signed-out visitors: expect the first 20 to 50 results per search reliably and deeper pages on a best-effort basis (the status message says when a search stopped early). Author articles go up to 5,000, because they are stored inside the profile row, which must stay under the platform's 10 MB item size.
• Abstracts are Scholar's snippets, not full abstracts.
• "Cite" formats (BibTeX, APA, MLA) and per-article citation graphs on profiles are not exported.
• Scholar sometimes throttles pages. The actor retries them; if a page still comes back short, the status message says so and you can rerun.