News Scraper: GDELT Articles, Full Text & 65 Languages

Search news from around the world by keyword and get back a clean list of articles: title, link, outlet, country, language, the time it was published online, and the share image. Filter by country, language and website, go back up to about 3 months, and get up to 5,000 articles per search. Turn on full text and it also visits each article and pulls out the story itself.

It searches the GDELT Project, an open index that monitors online news in more than 65 languages and is updated every 15 minutes. No Google pages are scraped, so it is a Google News alternative built on a source that is free to use and redistribute with credit.

Who it's for

• Media monitoring and PR: track mentions of a brand, a person or a topic across thousands of outlets, every day.
• Research and data teams: build news datasets by country and language for analysis or model training.
• Newsletters, alerts and dashboards: schedule a run every hour and feed new articles into Sheets, Slack, Make, Zapier or n8n.
• AI and RAG pipelines: get the article text, not just the headline.

How it compares

	This Actor	Google News scrapers on the Store
Source	GDELT DOC 2.0 API: open data, free to use with credit	Google News pages
Price per 1,000 articles	$1.50 (full text +$2.00, only when extracted)	$0.87 to $4.00
Articles per search	up to 5,000 (GDELT's 250-per-request cap is paged for you)	varies by tool
Filters	country, 65 languages, website, date range, GDELT operators (themes, tone, proximity)	varies by tool
Full article text	optional; robots.txt obeyed, paywalls left alone	varies by tool
Empty or failed searches	free	varies by tool

Store prices as seen in November 2026; they may change.

Input

Field	What it does
Search terms	One search per line. All words must appear. "exact phrase", (a OR b) and -word work, and so do GDELT operators such as sourcelang:spanish, sourcecountry:UK, theme:ENV_CLIMATECHANGE, tone<-5 and near20:"grain exports".
Time span	How far back from now: 15min, 24h, 7d, 2w, 3m (default 7d, at most 3m).
Start date / End date	A fixed range instead (2026-10-01, or 2026-10-01T08:00:00Z).
Max articles per search	1 to 5,000 (default 250).
Sort by	Newest first (default), oldest first, most relevant, most positive or most negative tone.
Publisher country	Country names (Germany, United Kingdom) or GDELT's FIPS codes (GM, UK). Note that FIPS codes are not ISO codes: GB is Gabon and AU is Austria in FIPS. Names are safest; the log flags codes that look like ISO codes.
Language	Spanish, Japanese or es, ja or spa, jpn.
Only these websites	e.g. reuters.com, bbc.co.uk (subdomains included).
Fetch full article text	Off by default. Adds text, textStatus and wordCount.
Skip articles found by an earlier search	On by default: an article that several searches find is saved and charged once.
Use Apify Proxy for GDELT requests	Off by default. In our tests it did not make runs faster, so leave it off unless a run keeps failing with rate-limit errors.

Example:

{
  "queries": ["\"electric vehicles\"", "inflation sourcecountry:UK", "election sourcelang:spanish"],
  "timespan": "7d",
  "maxResultsPerQuery": 500,
  "sort": "newest",
  "fetchArticleText": false
}

Output

One row per article (values illustrative):

{
  "query": "\"electric vehicles\"",
  "status": "found",
  "title": "EU electric vehicle sales climb for a third month",
  "url": "https://www.example-news.com/business/eu-ev-sales-climb",
  "domain": "example-news.com",
  "sourceCountry": "United Kingdom",
  "language": "English",
  "seenDate": "2026-11-03T09:15:00Z",
  "socialImage": "https://www.example-news.com/images/ev-sales.jpg",
  "mobileUrl": null,
  "text": "EU electric vehicle sales climbed for a third month in October...",
  "textStatus": "ok",
  "wordCount": 642,
  "textError": null,
  "error": null,
  "scrapedAt": "2026-11-03T09:30:12+00:00"
}

• seenDate is when GDELT first saw the article online (UTC). GDELT checks for news every 15 minutes, so this is close to, but not always the same as, the publication time.
• text, textStatus, wordCount and textError appear only when Fetch full article text is on. textStatus is one of ok, paywalled, blocked_by_robots (the publisher's robots.txt asks tools not to read it), blocked (the site refused, e.g. HTTP 403), not_found, http_error, unreachable, not_html, no_text_found (for example a page that needs JavaScript or a consent click) and not_public_address. Only ok rows carry text.
• A search with no articles gets one row with status: "no_results", and a search GDELT rejected or could not answer gets one row with status: "query_failed" and GDELT's reason in error. Both are free.
• The run also saves a SUMMARY record: counts per search, the exact GDELT query each search became, articles and texts charged, GDELT requests and rate-limit waits, and whether the spending cap stopped the run.

Pricing

Pay per event, no start fee:

• $0.0015 per article found ($1.50 per 1,000 unique articles).
• $0.002 per full text extracted ($2.00 per 1,000), charged only when text was actually extracted. Paywalled, blocked or empty pages cost nothing extra.

So 1,000 articles cost $1.50, or at most $3.50 with full text. Searches with no results and failed searches are free. Set a maximum charge per run in the run options and the Actor stops cleanly when it's reached, before fetching anything it couldn't charge for.

Good to know

• About 3 months back. GDELT's search covers a rolling window of roughly the last 3 months. For older news, use a dated archive.
• Speed. GDELT asks for no more than one request every 5 seconds, and each request returns at most 250 articles. A 1,000-article search takes about half a minute; ten searches of 250 take about a minute. When GDELT is busy it asks callers to slow down; the Actor waits and retries (10 s, 20 s, 40 s, 60 s, 60 s) and says so in the log. These waits are common from shared cloud machines: in our tests on Apify, 3 searches with full text took about 3 minutes. Searches that still fail after the retries are saved as free query_failed rows.
• Past 250 articles the date range is walked newest to oldest (or the reverse), so results stay in order with no repeats. For the relevance and tone sorts the range is split into time windows and the windows are interleaved, so the order is "best of each window" rather than one global ranking.
• Tone can be used to filter (tone<-5 in a search) and to sort, but GDELT's article list doesn't include a tone score per article, so there is no tone column.
• Languages. GDELT machine-translates the languages it monitors, so English search terms also find articles in other languages. Titles are kept in their original language.
• Coverage is what GDELT monitors: very broad, but not every outlet, and a few links may have moved or gone by the time you open them.
• Full text is read like a polite visitor would: one page per second per website at most, the publisher's robots.txt obeyed for every page and redirect, at most 3 MB per page, and no logins, cookies, paywall tricks or headless browser. Expect text for most open news sites and none for paywalled ones.

Source and credit

Article data comes from the GDELT Project (https://www.gdeltproject.org/) through its DOC 2.0 API. GDELT's data is free for any use, including commercial use, with credit to the GDELT Project and a link to its website; please keep that credit when you publish data from this Actor. The article texts belong to their publishers: check their terms before republishing them.

This Actor isn't affiliated with or endorsed by the GDELT Project, Google or any publisher.

More tools from the same developer

• RSS Feed Reader: Atom & RSS to JSON with Feed Finder
• Hacker News Scraper: Search, Stories & Comments
• Google Trends Scraper: Interest Over Time, Regions & Queries
• OpenAlex Papers Search: Scholarly Works, Citations & Abstracts
• ClinicalTrials.gov Search: Studies, Phases, Sponsors & Status
• Federal Contracts Search: SAM.gov Opportunities & Award History