Wikipedia Scraper - Articles & Content for AI / RAG

This Wikipedia scraper turns the world's largest encyclopedia into clean, structured data rows — search by keyword or fetch exact article titles, in any of Wikipedia's 300+ language editions, and get one flat row per article: title, URL, an intro summary or the full plaintext content, word count, categories, thumbnail, and last-edited date. No HTML soup, no wiki markup, no [edit] links — just clean text ready to drop into a vector store, a fine-tuning set, or a spreadsheet.

Built for AI / RAG pipelines, LLM training datasets, researchers, and content teams who need Wikipedia content as data — straight from the official MediaWiki API, no API key, no anti-bot, no proxies.

---

Why use it

• Two modes: search or exact titles — cast a wide net with a searchQuery ("artificial intelligence" returns the top matching articles), or pass a list of exact titles you already know you want. Both in one actor.
• Summary or full content — flip fullContent off for a tight intro summary (great for cards, previews, embeddings of short chunks), or on for the entire article as clean plaintext (great for chunking into a RAG store or a training corpus).
• Clean plaintext, not HTML — the MediaWiki explaintext extract strips markup, tables, and reference clutter, so you get readable prose with no post-processing.
• Any language edition — set language to en, es, fr, de, ar, zh, or any of Wikipedia's 300+ editions to build multilingual datasets.
• Rich metadata per row — word_count, up to 10 categories, a thumbnail_url, last_edited timestamp, and the canonical article url for attribution.
• Batched + resilient — content is fetched in batches with retry/backoff; if Wikipedia is briefly unavailable the run exits cleanly with a status message and you're not charged for an empty run.
• No API key, no proxies — the MediaWiki API is free and keyless. It runs fine on datacenter IPs.

---

What it does

You give it a search term (or a list of titles) and a language. The actor queries the MediaWiki search API to find the top matching articles, then pulls each article's content and metadata in batches of 20. Every article is flattened into one clean row and pushed to the dataset. Export to CSV, JSON, or Excel, or pull it straight into your pipeline via the API.

Use cases

• RAG / retrieval systems — build a knowledge base for a chatbot or agent. Pull full-content articles on a topic, chunk them, embed them, and ground your LLM in factual, citable sources with a real URL for each.
• LLM training / fine-tuning data — assemble a clean, deduplicated text corpus on any domain or in any language without scraping raw HTML or parsing wiki markup.
• Research & literature scans — grab the top 100 articles on a subject and get a word-count-ranked overview with summaries and categories in minutes.
• Content aggregation & enrichment — pull authoritative summaries, thumbnails, and categories to enrich a product catalog, a glossary, a news app, or an internal wiki.
• Multilingual datasets — run the same query across en, es, fr, and more to build parallel or comparative corpora.

---

How to use it

1. Enter a search query (e.g. artificial intelligence) — or leave it and provide a list of exact titles instead.
2. Set the language edition (default en).
3. Turn full article content on for the whole article text, or leave it off for the intro summary.
4. Set max articles and run → get a clean, deduped article dataset.

Input

Field	Type	Description
searchQuery	string	A search term; returns the top matching articles. Use this or titles.
titles	array	A list of exact article titles to fetch (e.g. ["Alan Turing", "Python (programming language)"]).
language	string	Wikipedia language edition code (en, es, fr, de, ar, zh, …). Default en.
fullContent	boolean	false (default) returns an intro summary; true returns the full content.
maxItems	integer	Max articles to return (1–500). Default 25.

You must provide at least one of searchQuery or titles.

Example input:

{
  "searchQuery": "artificial intelligence",
  "language": "en",
  "maxItems": 25,
  "fullContent": false
}

Output fields

Field	Description
title	Article title
pageid	Wikipedia page ID (stable identifier)
url	Canonical article URL (use for attribution)
summary	Intro extract — present when fullContent is false
content	Full article plaintext — present when fullContent is true
word_count	Word count of the returned text
categories	Up to 10 article categories
thumbnail_url	Lead image thumbnail URL (if the article has one)
last_edited	Timestamp of the last edit (touched)
language	Language edition of the article

JSON output sample

{
  "title": "Artificial intelligence",
  "pageid": 1164,
  "url": "https://en.wikipedia.org/wiki/Artificial_intelligence",
  "summary": "Artificial intelligence (AI) is the capability of computational systems to perform tasks typically associated with human intelligence, such as learning, reasoning, problem-solving, perception, and decision-making...",
  "word_count": 142,
  "categories": [
    "Category:Artificial intelligence",
    "Category:Cybernetics",
    "Category:Computational fields of study"
  ],
  "thumbnail_url": null,
  "last_edited": "2026-07-07T06:31:50Z",
  "language": "en"
}

Results render as a clean, sortable table on the Output tab and export to CSV, JSON, or Excel.

Example output

A real sample from a live run (searchQuery: "artificial intelligence", en):

Title	pageid	Words	URL
Artificial intelligence	1164	142	https://en.wikipedia.org/wiki/Artificial_intelligence
Artificial general intelligence	586357	118	https://en.wikipedia.org/wiki/Artificial_general_intelligence
A.I. Artificial Intelligence	142224	96	https://en.wikipedia.org/wiki/A.I._Artificial_Intelligence

---

Use with AI agents & automation

Run from the Apify MCP server so AI agents (Claude, ChatGPT, Cursor) can pull Wikipedia articles as a tool call, schedule runs via Make, n8n, or Zapier to feed a vector database, or sync the dataset to Google Sheets. Clean flat JSON with a stable pageid and a source url drops into RAG and dataset pipelines with no glue code.

---

Pricing

Pay-per-event — charged per article delivered. The source data is the free public MediaWiki API (no proxy or third-party cost), so you only pay for the clean, structured rows you actually receive — no subscription, no charge for empty runs. New Apify accounts get free platform credits each month, so you can pull and test a full topic before paying anything. See the Apify Store page for the current per-result price.

---

FAQ

Where does the data come from? The official MediaWiki API that powers Wikipedia — the same API Wikipedia's own apps use. It's public, keyless, and returns official article content.

Do I need an API key? No. It works key-free, with no proxies.

Is this the full article or just the intro? Your choice. Leave fullContent off for the intro summary; turn it on for the entire article as clean plaintext.

What about licensing and attribution? Wikipedia article text is released under CC BY-SA (Creative Commons Attribution-ShareAlike). You're free to reuse it — including for AI/RAG and commercial use — as long as you attribute Wikipedia and share adaptations under the same license. Each row includes the article url so you can attribute the source. Review the Wikipedia reuse guidelines for your specific use.

Which languages are supported? All of Wikipedia's 300+ editions — just set the language code (en, es, fr, de, ja, ar, zh, …).

Why did an exact title return nothing? Titles are matched exactly (capitalization is auto-corrected, but spelling and disambiguation like (programming language) must match). If unsure, use a searchQuery instead and let the search find the article.

Can I export to CSV or Google Sheets? Yes — CSV, JSON, or Excel from the Output tab, or sync to Google Sheets via Make, n8n, or Zapier.

---

Other Flash Scrape scrapers

• AI Web Scraper — extract structured data from any web page with AI
• Google SERP Scraper — Google search results as clean rows

Questions or a field you need added? Reach out via the Issues tab on the Actor's Apify Store page — happy to help.