GDELT News Monitor: keyword alerts, tone, themes, people and organizations

Monitor news from around the world by keyword, source domain, country and language, straight from the GDELT Project's 15-minute article feed. Every row is one news article with its URL, title, source domain, tone score, GDELT themes, mentioned locations and organizations (people only if you switch them on). Watch mode returns only the articles that appeared since your last run, so a scheduled run becomes an alert feed.

The feed covers online news in 65+ languages. This Actor reads GDELT's published files directly (no scraping of news sites, no proxies, no search-engine results), so it is stable and cheap to run. Speed depends on how much of the feed you read: a watch-mode run (a window of 15 to 60 minutes) finishes in seconds, a 6-hour run in under a minute, and a full 24-hour run across all languages reads about 760 MB of files and can take around 10 minutes. Set includeTranslated to false, or filter by languages, to read English-language sources only, which is about a third of the data.

What you get

• Filters: keywords (whole-word, any of, plus an exclude list), source domains, countries mentioned, source languages, tone range, a time window up to 7 days back.
• Per article: url, title, sourceDomain, language, seenAt, publishedAt (when the page declares it), tone, positiveScore, negativeScore, polarity, wordCount, themes, locations (name, country code, latitude, longitude), organizations, imageUrl, matchedKeywords, and persons only when you switch on includePersons.
• Watch mode: the position is saved in a named key-value store on your account (gdelt-news-monitor-state). The next run continues exactly where the last one ended, and articles are never returned twice.

Input example

{
  "keywords": ["Nvidia", "semiconductor"],
  "excludeKeywords": ["horoscope"],
  "languages": ["en"],
  "hoursBack": 24,
  "watchMode": true,
  "maxItems": 500
}

Leave every filter empty to run a free-form 5-article example (keyword election, last two hours).

Output example

{
  "url": "https://www.example-news.com/nvidia-earnings-preview",
  "title": "Nvidia earnings preview: what analysts expect",
  "sourceDomain": "example-news.com",
  "language": "eng",
  "translated": false,
  "publishedAt": "2026-11-01T13:02:11Z",
  "seenAt": "2026-11-01T13:45:00Z",
  "tone": -1.42,
  "positiveScore": 2.1,
  "negativeScore": 3.5,
  "polarity": 5.6,
  "wordCount": 612,
  "themes": ["ECON_STOCKMARKET", "TAX_FNCACT_ANALYST", "WB_1920_FINANCIAL_SECTOR_DEVELOPMENT"],
  "locations": [{"name": "Santa Clara, California, United States", "countryCode": "US", "lat": 37.35, "lon": -121.95}],
  "persons": [],
  "organizations": ["nvidia", "securities and exchange commission"],
  "matchedKeywords": ["nvidia"],
  "source": "GDELT Project",
  "sourceUrl": "https://www.gdeltproject.org/",
  "citation": "Data: The GDELT Project (https://www.gdeltproject.org/). Any use or redistribution of GDELT data must include a citation to the GDELT Project and a link to https://www.gdeltproject.org/.",
  "scrapedAt": "2026-11-01T14:00:00Z"
}

Pricing

$0.002 per article returned, with no start fee. Reading files and filtering is free: a run that finds nothing costs nothing.

• 100 articles: $0.20
• 1,000 articles: $2.00
• A watcher on a narrow keyword (about 20 new articles an hour): roughly $1 a day

Set maxItems (or the run's maximum cost) to cap a run.

Citation and licence

All article data comes from The GDELT Project (https://www.gdeltproject.org/), which makes it available for unrestricted academic, commercial and governmental use. GDELT's terms require that any use or redistribution of the data includes a citation to the GDELT Project and a link to https://www.gdeltproject.org/. Every row carries that citation in the citation field; keep it when you republish or share the data. This Actor is not affiliated with or endorsed by GDELT.

How it differs from a news-search scraper

• The source is GDELT's open event feed, not a search engine. Results are not ranked by relevance or popularity: you get every matching article GDELT saw in the window, in time order.
• Rich entities and tone for every article: themes, places with coordinates, people, organizations and a tone score, which plain headline scrapers do not return.
• Open licence. The data may be stored and redistributed with the citation above.

Limits and good to know

• Titles and matching. Keywords are matched against the title, URL and organizations of each article (and the people, when includePersons is on), not the full text. About 98% of articles carry a title.
• Non-English articles come from GDELT's machine-translated stream: the title stays in the original language, so match those with keywords in that language or filter by countries. Themes and organizations are normalized to English.
• Countries filters by the places an article mentions. GDELT does not record where the publisher is based.
• Latency: GDELT publishes a new file every 15 minutes, a few minutes after each quarter hour, so alerts lag the original story by roughly 15 to 30 minutes.
• Window: up to 168 hours (7 days) per run. Files GDELT did not publish are skipped and reported in the status message.
• Not every news site is monitored by GDELT. Coverage is broad, not complete.

FAQ

How do I get alerts? Turn on watchMode, then schedule the Actor (Apify Schedules) every 15 to 60 minutes and connect the dataset to a webhook, Zapier, Make or n8n.

Can two watchers use the same keywords? Yes, give each a different watchName.

Why is publishedAt often empty? It is filled only when the page itself declares a publication time. seenAt is always present: it is when GDELT processed the article.

Does it return the article text? No. It returns the URL, title and GDELT's analysis. Fetch the text separately from the publisher where its terms allow.

Purpose limit

This Actor is for monitoring topics, companies, organizations and public figures in the news. Do not use it to track, profile or contact private individuals by name. News articles also name victims, suspects, minors and bystanders, so includePersons is off by default; turn it on only for a public-figure watchlist, and do not use the output for credit, employment, insurance, housing or similar decisions about a person.

Removal requests

If your name appears in a persons list and you want it dropped, email datagleaner@gmail.com with the name as it appears. We answer within 30 days, record each request with its date and answer, and add the name to the Actor's exclusion list in the next release. The Actor keeps no copy of the data: the articles stay with their publishers and the GDELT Project.

Responsible use

The data is GDELT's public analysis of published news. Person and organization names are those that appear in public articles. Use the data lawfully, keep the GDELT citation, and respect the terms of the original publishers when you open their pages.