News Search - Global News by Keyword, Google News Alternative

Search news articles from the last 72 hours worldwide by keyword, source domain, country, language or theme, using the GDELT Project's open 15-minute feeds. Each article: URL, title, source, time, tone, themes, locations, organisations and image. No API key. $1.20 per 1,000 articles.

A news search API for the whole world's news, without Google News scraping or an API key. Give it keywords (a company, a product, a topic, a person in the news) and it returns matching news articles from the last 1 to 72 hours, newest first: article URL, page title, source domain, the time the article was seen (15-minute resolution), language, tone and sentiment, GDELT themes, locations and countries mentioned, organisations, lead image and which keywords matched. Filter by source domain (e.g. only reuters.com and bbc.co.uk, or exclude sites), country mentioned, language (English plus about 65 translated languages: German, French, Spanish, Chinese, Arabic, Russian, Japanese and more), GDELT theme and sentiment. Switch on monitor mode for scheduled runs and each run returns only articles newer than the last one, which makes it a cheap news monitoring and brand mention feed. Use it for media monitoring, PR and brand tracking, market and competitor news, financial news signals, news datasets for AI agents and RAG pipelines, and as a Google News alternative that works from any server. Chain it with Website to Markdown when you need the full article text. The data comes from the open GDELT Project 15-minute GKG files, which GDELT licenses for unlimited commercial use with attribution; every row carries a source credit. $1.20 per 1,000 articles, with no compute bill on top. Runs that find nothing return a free status row instead of failing.

What data does it return?

Every run outputs clean, schema-validated rows like these:

publishedAt	title	sourceDomain	language	sentiment	tone	countries	matchedKeywords
2026-10-23T10:00:00.000Z	OpenAI partners with Poland's NASK to strengthen AI safety	thenews.com.pk	en	positive	2.68	United Kingdom, Poland	artificial intelligence
2026-10-23T10:00:00.000Z	World shares are mixed after Wall Street's AI-related losses, while oil prices jump	union-bulletin.com	en	negative	-3.55	Australia, France, Japan, United Kingdom, …	Nvidia
2026-10-23T08:30:00.000Z	Ohne Lenkrad und Pedale: Teslas neues Robotaxi soll noch 2026 in den Verkauf starten	az-online.de	de	neutral	0.98	Germany	Tesla

How much does it cost?

Transparent pay-per-event pricing — you only pay for what you get, and the actor always respects your run budget: set a maximum total charge and it stops cleanly at the cap instead of overrunning.

Event	Price	Charged	Worked example
article	$0.0012	per news article row	500 articles about your brand = $0.60
apify-actor-start	$0.00005	per run per GB of memory (platform start fee)	One run at 1 GB = $0.00005

Input options

Field	Type	Required	Default	Description
keywords	array	no	—	Words or phrases to find, e.g. "artificial intelligence", "Tesla" or "interest rates". Matched as whole words, case-insensitive, in the article title, its URL, the organisations GDELT extracted and the names mentioned in the article (not the full article text). Leave empty to get all articles that pass the other filters.
keywordMatch	string	no	"any"	Any: an article matching at least one keyword is returned. All: every keyword must match.
excludeKeywords	array	no	—	Drop articles that mention any of these words or phrases (same matching as keywords).
lookbackHours	integer	no	24	How far back to search, from the newest GDELT batch: 1 to 72 hours. The translated (non-English) feed is searched over at most 24 hours. Newest articles come first.
maxArticles	integer	no	100	Stop after this many articles (newest first). Each article is one charged row.
feed	string	no	"english"	English: English-language news worldwide. Translated: news in about 65 other languages (titles stay in the original language; organisations and themes are in English). All: both. A non-English language filter switches to the translated feed automatically.
languages	array	no	—	Only articles in these languages, as ISO codes like en, de, fr, es, zh, ar (two- or three-letter). Empty = all languages in the chosen feed.
countries	array	no	—	Only articles that mention a place in one of these countries, by English country name, e.g. "Germany", "United States", "United Kingdom", "India". Case-insensitive.
sourceDomains	array	no	—	Only articles from these news sites, e.g. bbc.co.uk, reuters.com, lemonde.fr. Subdomains are included.
excludeSourceDomains	array	no	—	Drop articles from these sites (subdomains included).
themes	array	no	—	Only articles tagged with one of these GDELT GKG themes. A full code (ECON_STOCKMARKET) or a prefix ending at an underscore boundary (ECON, TAX_FNCACT, WB_). See the GDELT GKG theme list.
sentiment	string	no	"any"	Only articles with this overall tone. GDELT tone ≥ 1.5 is positive, ≤ −1.5 negative, otherwise neutral.
requireTitle	boolean	no	true	Skip the rare articles for which GDELT recorded no page title.
includePeople	boolean	no	false	Add the names of people GDELT found in each article (mostly public figures). Off by default; not intended for profiling individuals.
newSinceLastRun	boolean	no	false	For scheduled runs: skip GDELT batches already scanned by an earlier run with the same filters, so each run returns only newer articles. State is kept in a key-value store named news-search-state in your account.

Limits & honest notes

• Data: the GDELT Project's open GKG 2.1 files (https://www.gdeltproject.org/), published every 15 minutes and licensed for unlimited use with attribution. Every row has source "GDELT Project (https://www.gdeltproject.org/)"; please keep that credit when you republish the data. Not affiliated with GDELT or Google.
• Keywords match whole words in the title, the URL, the organisations GDELT extracted and the names mentioned in the article. They are not matched against the full article text, so recall is lower than a full-text search engine. Use several related keywords (e.g. "artificial intelligence" and "AI") for better coverage.
• publishedAt is the 15-minute GDELT batch in which the article was first seen (usually within 15 minutes of publication), not the publisher's own timestamp. publishedAtPrecision says so on every row.
• Window: up to 72 hours back for English-language news, up to 24 hours for the translated feed (about 65 languages; titles stay in the original language). GDELT sometimes skips or delays a batch; those are reported as missing in the run summary and skipped.
• Coverage is what GDELT monitors: tens of thousands of news sites worldwide, roughly 1,000 English and 3,000 translated articles per 15 minutes. There is no "top stories" ranking: results are newest first.
• No article bodies. Chain Website to Markdown on the URLs if you need full text.
• Countries are matched by English country name in GDELT's location data ("Germany", "United Kingdom"); the locations field also gives GDELT's FIPS country codes and coordinates.
• People names are off by default. Switching on includePeople adds names of people mentioned (mostly public figures); do not use them to profile individuals.
• Respects your run budget: set a maximum total charge and the run stops cleanly at the cap, with nothing charged twice. A run with no matches returns one free status row explaining why.

Related ActorForge actors

• Website to Markdown — full article text for the URLs found here
• RSS Feed Reader — follow specific news sites' own feeds

Changelog

• 2026-10-23 — Initial release: keyword news search over GDELT's 15-minute GKG files (English and translated feeds), filters for language, country, source domain, theme and sentiment, URL de-duplication, monitor mode, free status rows.

---

Built and maintained by ActorForge. Reliability first: every actor in the portfolio is monitored by an automated nightly self-test, and breakages are typically fixed the same day.