Google News Scraper API — Full Text, Monitoring & RSS

Search Google News by keyword and get back the publisher's real link, the full article text, and on repeat runs only the stories you haven't seen yet. No API key, no login. You can fold in the GDELT global news index and any RSS or Atom feed as well, and it all arrives as one table you can export to JSON, CSV or Excel.

What does Google News Scraper API do?

• Searches Google News and gives you the publisher's own link, not a news.google.com/rss/articles/CBMi… redirect
• Pulls the article body, so you get more than a headline and two lines of summary
• Goes past Google's ceiling of roughly 100 articles per query by walking a date range one day at a time
• Remembers what it already sent you, so a scheduled run reports only new stories
• Filters on publisher domain and on whole words, so car never matches cargo
• Narrows to a language or a country from a dropdown, covering 146 languages and 235 countries
• Reads GDELT and any RSS feed into the same table, with BBC, CNN, NPR, New York Times, Guardian, Google News and Hacker News ready to pick from

What people use it for

• Watching for mentions of a brand or a competitor, with one row per story instead of the same piece three times
• Media monitoring and PR reporting, where you need the publisher, the date and a link that stays stable
• Building research datasets, with the article text across a stretch of dates
• Feeding alerts: schedule it, get only what's new, send it to Slack or a webhook

How to scrape Google News

1. Click Try for free to open the Actor in Apify Console
2. Type a keyword into Search queries, say tesla recall
3. Choose your Sources: google-news, gdelt, rss, or all three
4. Add Preset feeds (BBC, Guardian, and so on) or paste feed URLs of your own, if you want them
5. Set Maximum results and hit Start
6. Grab the results from the Dataset tab as JSON, CSV or Excel

Three switches are worth knowing about. Include full text fetches each article and pulls out the body. Resolve Google News URLs turns those redirect links into real publisher links, and full text turns it on for you anyway. Skip already seen articles is the one that makes scheduled runs useful, because each run then brings back only what the last one missed.

Use it as a Google News API

A single request gives you the rows straight back, so you can treat it as a news endpoint from any language:

curl -X POST "https://api.apify.com/v2/acts/automly~google-news-scraper-api/run-sync-get-dataset-items?token=YOUR_API_TOKEN" \
  -H 'Content-Type: application/json' \
  -d '{
    "queries": ["tesla recall"],
    "sources": ["google-news", "gdelt"],
    "presetFeeds": ["bbc", "guardian"],
    "maxResults": 50,
    "includeKeywords": ["tesla"],
    "resolveGoogleUrls": true
  }'

Or from Python, using the apify-client package:

from apify_client import ApifyClient

client = ApifyClient("YOUR_API_TOKEN")
run = client.actor("automly/google-news-scraper-api").call(
    run_input={"queries": ["tesla recall"], "includeFullText": True, "maxResults": 50}
)
for article in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(article["publishedAt"], article["sourceDomain"], article["title"])

You can also schedule the same run in Apify Console and wire it up to Slack, Google Sheets, webhooks, Make, n8n or Zapier through Apify integrations.

Input parameters

Parameter	Type	Description
queries	string[]	Search terms, one per line, up to 100. Sent to every search source you enabled
feedUrls	string[]	RSS or Atom feed URLs to read directly, up to 200
presetFeeds	string[]	Well-known feeds by name: bbc, cnn, npr, nyt, guardian, google-news-top, hacker-news
sources	string[]	Any of google-news, gdelt, rss. Defaults to google-news
maxResults	integer	How many articles to collect in total, 1–50000. Defaults to 100
sinceHours	integer	Only articles from the last N hours, 1–8760
fromDate / toDate	string	YYYY-MM-DD bounds. A date range wins over sinceHours
includeKeywords	string[]	Keep articles whose headline or summary contains one of these. Whole words, case-insensitive. Filtering happens before the body is fetched, so bodies aren't searched
excludeKeywords	string[]	Drop articles containing any of these. Exclusions beat includes
includeDomains	string[]	Keep only these domains, like reuters.com. Subdomains count, and www. and ports are ignored
excludeDomains	string[]	Drop these domains and anything under them
language	string	Pick a language from the dropdown. It sets the Google News edition and drops rows from sources that report a different language
country	string	Pick a country from the dropdown. It sets the Google News edition and drops rows from sources that report a different country
resolveGoogleUrls	boolean	Turn Google News redirect links into real publisher links
includeFullText	boolean	Fetch each article page and pull out the body. Resolves Google News links for you. Some publishers refuse automated fetches, so a few rows keep their summary and leave the body empty
skipSeenArticles	boolean	Return only articles that earlier runs didn't deliver
seenArticlesStoreName	string	Names the store of delivered articles. Use a separate one per monitored query
proxyConfiguration	object	Proxy settings. Leave a rotating group switched on

You need at least one of queries, feedUrls or presetFeeds. Adding a feed or a preset switches the rss source on by itself.

Output example

One row per article, nineteen fields. This is a real row from a run on 12 September 2026:

{
  "title": "Lawmaker Demand Feds Take Action After 43 Tesla Drivers Filmed Asleep At The Wheel",
  "url": "https://insideevs.com:443/news/807893/tesla-sleeping-congress-nhtsa-demand/",
  "canonicalUrl": "https://insideevs.com/news/807893/tesla-sleeping-congress-nhtsa-demand",
  "source": "gdelt",
  "sourceName": "insideevs.com",
  "sourceDomain": "insideevs.com",
  "publishedAt": "2026-09-11T23:15:00+00:00",
  "summary": null,
  "author": null,
  "imageUrl": "https://cdn.motor1.com/images/mgl/W87NGL/s1/tesla-fsd-with-a-dead-battery.jpg",
  "categories": [],
  "language": "English",
  "country": "United States",
  "query": "tesla recall",
  "guid": null,
  "resolved": true,
  "fullText": null,
  "fullTextChars": null,
  "scrapedAt": "2026-09-12T12:52:24.373597+00:00"
}

canonicalUrl is the field deduplication works from. It lowercases the host, drops www., the fragment and a redundant port, and strips the usual tracking parameters such as utm_*, at_*, fbclid and gclid. That's why the same story reaching you from two different feeds still counts once.

Switch on resolveGoogleUrls and a Google News row carries the publisher's own link instead:

{
  "title": "Driver crashes Tesla into scaffolding in midtown Manhattan, killing passenger: Police - ABC News",
  "url": "https://abcnews.com/US/tesla-crashes-scaffolding-midtown-manhattan-killing-passenger/story?id=136300943",
  "source": "google-news",
  "sourceDomain": "abcnews.com",
  "resolved": true
}

Switch on includeFullText and each row gains the body:

{
  "title": "Tesla and others begin record vehicle recall in China - Reuters",
  "url": "https://www.reuters.com/world/tesla-fix-software-millions-china-made-imported-evs-china-2026-08",
  "sourceDomain": "reuters.com",
  "fullTextChars": 4258,
  "fullText": "BEIJING, Aug 21 (Reuters) - Tesla and eight other automakers said on Friday they will recall a total of about 4.3 million vehicles in China over concerns that doors may be difficult to open…"
}

The three sources don't publish the same things, so some columns are fuller than others:

Source	You get	You don't
rss	headline, link, date, summary, guid, and author, image or categories when the feed bothers to include them	language, country
google-news	headline, link, date, summary, guid, publisher domain	author, categories, language, country
gdelt	headline, link, date, image, language, country	summary, author, guid

fullText and fullTextChars only appear when you ask for full text.

How many Google News articles can you get per query?

Google News gives you about 100 articles for a single query, and GDELT stops at 250. To go further, ask for more than that and give a date range. The range then gets split into daily windows and queried a day at a time, so a week of history can return several hundred rather than one hundred. Ask for 500 with no date range and you'll still hit the ceiling.

When you pick several sources, rows come back round-robin rather than one source at a time. Ask for 60 across three sources and you'll get roughly 20 from each, instead of 60 from whichever happened to be fetched first.

Monitoring: only new articles on every run

Switch on skipSeenArticles and every run returns only the articles no earlier run delivered, which is what makes polling on a schedule worthwhile. In testing, a first run stored 60 articles and an identical second run stored 34. Give each query you monitor its own seenArticlesStoreName so they don't share a memory.

What it doesn't do

No sentiment scoring, no entity extraction, no translation, no paywall bypass. GDELT also rate-limits by IP and will refuse a request now and then, which shows up in the log as GDELT returned HTTP 429. The run retries on a fresh address, and your other sources carry on unaffected.

FAQ

Do I need an API key for Google News or GDELT?

No. It reads Google News's public RSS feeds and GDELT's public DOC API. Neither asks for credentials, and no login or cookies are involved.

Can I get more than 100 articles from one Google News query?

Yes. Set maxResults above the ceiling and give it fromDate and toDate. The range gets split into daily windows and queried one day at a time, so each day can add up to about another hundred articles.

Why do Google News links point at google.com instead of the publisher?

Because Google News publishes encoded redirect links, and following one lands you back inside Google rather than on the article. Switch on resolveGoogleUrls and every link becomes the publisher's own. That also lets the same story be matched against GDELT and RSS copies when duplicates are removed.

How do I get the full article text and not just the summary?

Switch on includeFullText. Each article page gets fetched and its main text pulled out — a Reuters piece came back at 4,258 characters — and Google News links are resolved along the way so the body is actually reachable.

Expect a few rows to come back without text. On one 80-article run, 71 had text and 9 didn't: those publishers either refused the fetch outright or served a page with no article in the HTML. Those rows keep their summary, and the body is left empty rather than stuffed with navigation text. The run log tells you the split, so you're not left guessing why a column has gaps.

Is the same article returned twice if two sources carry it?

Not once Google News links are resolved. Deduplication works from canonicalUrl, so one story counts once across all three sources. The exception is a Google News row you left unresolved: it has no publisher link to compare against, so it's keyed on its Google id and can't be matched with a GDELT or RSS copy of the same piece. Switching on resolveGoogleUrls brings those rows in too.

Can I use it instead of Google Alerts?

For keyword monitoring, yes. Schedule it, switch on skipSeenArticles, and connect the run to Slack, email or a webhook through Apify integrations. Each run then delivers only the stories earlier runs didn't, with the publisher link, the date and optionally the full text, ready to filter or export.

Is it legal to scrape Google News?

It collects publicly available article metadata from public RSS feeds and GDELT's public API, without logging in, and full text comes from the publisher's own public page. Complying with the terms of the sources you use, with copyright, and with data-protection law where you are is down to you.

What output formats are supported?

JSON, CSV, Excel, XML and JSONL, from the Dataset tab or the API, plus Apify integrations for Google Sheets, Slack, webhooks and S3.