Google News Scraper & API - News Monitoring & Real Article URLs

Google News scraper and unofficial Google News API: search Google News by keyword, read Google News topics (Top stories, World, Business, Technology...) or local news by city, and get the articles as JSON, CSV or Excel with headline, source, publish date, Google News link, the real publisher URL (optional link decoding) and related coverage of the same story. Works for any language and country edition, with news search time filters from the past hour to the past year, exact date ranges for historical news, and site filters like reuters.com.

It reads Google News' public RSS feeds over plain HTTP, with no headless browser: 100 articles arrive in about 2 seconds and 1,000 in under 10. Use it for news monitoring, brand and competitor tracking, PR and media analysis, stock and crypto news feeds, and feeding AI agents and RAG pipelines with fresh news.

What you get

One dataset item per article:

• Article: title (the outlet name Google appends is removed), publishedAt (ISO 8601, UTC), articleId (Google's stable ID, use it to deduplicate between runs).
• Source: sourceName ("Reuters"), sourceUrl ("https://www.reuters.com").
• Links: googleNewsUrl (always), publisherUrl (the decoded address on the publisher's site, when Resolve real publisher URLs is on), urlResolved (true / false), and url, which is the publisher URL when it was decoded and the Google link otherwise, so url is always usable.
• Related coverage: relatedArticles - other outlets' articles on the same story (title, sourceName, googleNewsUrl) and relatedCount. Topic, top-story and location feeds carry these on most rows (56 of 65 Technology rows in a test run); keyword searches rarely do.
• Context: position (order saved per query), feedType (search, topic, top, location), query, topic, location, searchQuery (the exact text sent to Google, with site: / when: / after: operators), language, country, feedUrl, scrapedAt.

A RUN_SUMMARY record in the key-value store lists per-query counts, feed requests made, how many feeds hit Google's 100-article limit, links resolved and why each query stopped.

Sample output

Real rows from test runs:

title	sourceName	publishedAt	feedType	query / topic	urlResolved
Our approach to EU text provenance rules	OpenAI	2026-10-05T15:12:28.000Z	search	openai	true
Fosi Audio Launches DAP Neo Digital Portable Music Player	Forbes	2026-10-10T04:00:00.000Z	topic	TECHNOLOGY	true
Trump really wants to make ‘super intelligence’ happen. Is it going to happen?	CNN	2026-10-11T16:00:44.000Z	search	artificial intelligence	false

Use cases

• News monitoring: schedule queries hourly with time range 1h, or daily with 1d, and only pay for the fresh articles.
• Brand, competitor and executive tracking: one query per name, plus site filters for the outlets you care about.
• Finance and crypto feeds: tickers and company names with 1h / 1d, pushed to Slack or a database.
• Media and PR analysis: count coverage by sourceName and day over a date range.
• Historical research: dateFrom / dateTo go back many years (a 2015 date range returned articles in testing).
• AI agents and RAG: one call returns structured, dated, sourced headlines; turn on link decoding and pass publisherUrl to a content extractor.

Pricing

$1.00 per 1,000 articles (pay per result; failed requests are never charged). The most-used Google News scrapers on the Store charge $2.00-$5.00 per 1,000, several add a start fee, and several charge $0.50-$1.50 per 1,000 extra to decode publisher URLs. Here link decoding costs nothing extra.

You scrape	You pay
100 articles	$0.10
1,000 articles	$1.00
10,000 articles	$10.00
100,000 articles	$100.00

No start fee. The Apify free plan's $5 monthly credit covers about 5,000 articles. The prefilled input (artificial intelligence, 20 articles) finishes in about 6 seconds and costs $0.02.

Why this one

• Cheapest of the most-used Google News scrapers: $1.00 per 1,000 with no start fee and free URL decoding, against $4.00 (data_xplorer), $2.30 + $0.005 per run (automation-lab), $2.00 + $0.005 per run + $1.00 per 1,000 decoded URLs (lhotanova) and $5.00 + $0.09 per run (easyapi).
• Real publisher URLs at no extra charge: optional decoding of Google's redirect links to the article on the publisher's site.
• Keyword, topic and city feeds in any language and country edition, in one actor.
• Time filters and date ranges: past hour to past year, or exact dateFrom / dateTo for historical news.
• Fast: public RSS over plain HTTP, 1,000 articles in under 10 seconds.

FAQ

Is it legal to scrape Google News? This actor collects only public headline listings: titles, source names, dates and links. No personal-data focus, no login, no article text. Google's RSS terms describe the feeds as meant for personal feed readers, and you're responsible for using the data in line with Google's terms, publishers' rights and applicable law. This is not legal advice.

Does it need a login, a browser or an API key? No. Plain HTTPS requests with normal browser headers to Google News' public RSS feeds.

How fast is it, and how many articles can I get? 100 articles take about 2 seconds, 1,000 under 10 (link decoding is rate-limited to roughly 5 articles per second). One Google News request returns at most about 100 articles; set a time window and a higher Max articles per query and the scraper searches day by day to go past 100 (tesla over 30 days reached 1,000 unique articles). See "How deep can it go?" below.

Can I schedule it or use it via API, and get only new articles? Yes. Schedule the run and set the time range to match (1h for hourly, 1d for daily); keep the articleId values you've seen for exact deduplication. Start runs from the Apify API, the JavaScript/Python clients, Make, Zapier, n8n or the Apify MCP server.

Do I need proxies? How does it avoid blocks? No setup needed: requests go through Apify Proxy (datacenter) by default with a session pool. A 429, 403, unusual-traffic or consent page retires that session and the request is retried on a fresh IP (6 retries by default). No blocks occurred in any test run.

Why are the links news.google.com/rss/articles/... and not the real article? That is what Google News publishes, and those links only redirect inside a browser. Turn on Resolve real publisher URLs to get the real address in publisherUrl and url. In testing, 2,033 of 2,033 links were decoded; each link gets 3 attempts on fresh IPs before the row is saved unresolved.

Which languages and countries work? Any Google News edition. Tested: en-US/US, en-GB/GB, de/DE, fr/FR, pt-BR/BR, es-419/MX, ja/JP, zh-TW/TW. The ceid parameter is built for you.

Example inputs for common jobs

Copy one into the JSON tab of the input and adjust.

Hourly brand or executive monitor (schedule it every hour; only the fresh articles are charged):

{ "queries": ["\"OpenAI\"", "\"Sam Altman\""], "timeRange": "1h", "maxItemsPerQuery": 100 }

Daily stock news feed with the real article links, ready for a content extractor or an LLM:

{ "queries": ["NVDA stock", "AAPL stock", "TSLA stock"], "timeRange": "1d", "resolveUrls": true }

Morning headlines digest for the UK edition (about 35-70 articles per section, about $0.15 for all three):

{ "topics": ["TOP", "BUSINESS", "TECHNOLOGY"], "language": "en-GB", "country": "GB" }

Local news for two cities (about 70 articles each):

{ "locations": ["Chicago", "Austin"] }

Media analysis: one quarter of coverage on a topic, searched day by day past Google's 100-article limit (up to 2,000 articles, $2.00 at most):

{ "queries": ["electric vehicles"], "dateFrom": "2026-07-01", "dateTo": "2026-09-30", "maxItemsPerQuery": 2000 }

What chosen outlets wrote about a competitor in the last week:

{ "queries": ["Rivian"], "sites": ["reuters.com", "bloomberg.com", "techcrunch.com"], "timeRange": "7d" }

How to use

1. Enter one or more Search queries (tesla, "climate change" -opinion, nvidia OR amd earnings), and/or pick Topics (Top stories, World, Nation, Business, Technology, Entertainment, Sports, Science, Health), and/or type Locations (Chicago, Sydney) for local news.
2. Set Language and Country (default en-US / US), a Time range (past hour, 12 hours, 24 hours, 7 days, 30 days, year) or Date from / Date to, and optionally Only these sites. Set Max articles per query (default 100).
3. Tick Resolve real publisher URLs if you need the article's own address instead of the news.google.com link.
4. Click Start. Download the results as JSON, CSV, Excel or HTML from the Output tab, or pull them through the Apify API, Make, Zapier, n8n or an MCP client.

Input example

{
    "queries": ["tesla", "openai", "nvidia earnings"],
    "topics": ["TECHNOLOGY"],
    "timeRange": "7d",
    "maxItemsPerQuery": 150,
    "resolveUrls": true
}

An exact date range in German, limited to two outlets:

{
    "queries": ["tesla"],
    "language": "de",
    "country": "DE",
    "dateFrom": "2026-09-01",
    "dateTo": "2026-09-05",
    "sites": ["spiegel.de", "handelsblatt.com"],
    "maxItemsPerQuery": 250
}

Everything two outlets published in the past 24 hours (no query needed):

{ "sites": ["reuters.com", "bbc.com"], "timeRange": "1d", "maxItemsPerQuery": 100 }

Output example

Real items from a cloud run of the first input above. A keyword search row:

{
    "position": 15,
    "articleId": "CBMiV0FVX3lxTFBndFg3dkRHQWJpbWRzMU1NVXZucE9POUFlMlNZemhnSlNOYTRNSDhEVFhnS3oxSVVEWFN5TDRjd3ZSTVJpdjFvUGQyc092YnVqX2ZxdEhCSQ",
    "title": "Our approach to EU text provenance rules",
    "sourceName": "OpenAI",
    "sourceUrl": "https://openai.com",
    "publishedAt": "2026-10-05T15:12:28.000Z",
    "url": "https://openai.com/index/eu-text-provenance/",
    "publisherUrl": "https://openai.com/index/eu-text-provenance/",
    "urlResolved": true,
    "googleNewsUrl": "https://news.google.com/rss/articles/CBMiV0FVX3lxTFBndFg3dkRHQWJpbWRzMU1NVXZucE9POUFlMlNZemhnSlNOYTRNSDhEVFhnS3oxSVVEWFN5TDRjd3ZSTVJpdjFvUGQyc092YnVqX2ZxdEhCSQ?oc=5",
    "relatedArticles": [],
    "relatedCount": 0,
    "feedType": "search",
    "query": "openai",
    "topic": null,
    "location": null,
    "searchQuery": "openai when:7d",
    "language": "en-US",
    "country": "US",
    "feedUrl": "https://news.google.com/rss/search?q=openai+when:7d&hl=en-US&gl=US&ceid=US:en",
    "scrapedAt": "2026-10-11T17:31:49.608Z"
}

A topic row with related coverage (Google links shortened here):

{
    "position": 32,
    "title": "Fosi Audio Launches DAP Neo Digital Portable Music Player",
    "sourceName": "Forbes",
    "sourceUrl": "https://www.forbes.com",
    "publishedAt": "2026-10-10T04:00:00.000Z",
    "url": "https://www.forbes.com/sites/marksparrow/2026/10/10/fosi-audio-launches-dap-neo-digital-portable-music-player/",
    "urlResolved": true,
    "relatedArticles": [
        { "title": "This Portable CD Player Evokes Discman Nostalgia With A MagSafe Twist", "sourceName": "SlashGear", "googleNewsUrl": "https://news.google.com/rss/articles/CBMiigFB...?oc=5" },
        { "title": "The Fosi Audio DISC Nano Is An Ultra-Portable CD Player", "sourceName": "Men's Gear", "googleNewsUrl": "https://news.google.com/rss/articles/CBMiVEFV...?oc=5" },
        { "title": "Fosi Audio DAP Neo Launches as a $79.99 Retro Music Player", "sourceName": "Techaeris", "googleNewsUrl": "https://news.google.com/rss/articles/CBMiekFV...?oc=5" }
    ],
    "relatedCount": 3,
    "feedType": "topic",
    "topic": "TECHNOLOGY",
    "language": "en-US",
    "country": "US"
}

How deep can it go? (read before large runs)

• One Google News request returns at most about 100 articles, and there is no page 2. That is a limit of Google News itself.
• To go past 100, set Max articles per query above 100 and a time window. The scraper first takes Google's ranked 100, then searches the window one day at a time (newest day first, up to 100 per day) and drops duplicates by article ID.
• Realistic depth: a busy query yields roughly 50-100 new articles per day this way. Measured: tesla over 30 days hit 1,000 unique articles after 14 day-slices; nvidia earnings over 7 days ended at 285. A niche query may have 5 a day. Days with more than 100 matching articles are capped at 100 (the RUN_SUMMARY shows how many feeds hit the limit); split such a query into narrower ones or add site filters to get more.
• Day slicing needs a window of at least two days: 7d, 30d, 1y or a date range. With anytime it walks back 30 days; with only Date to set it walks back 30 days from that date. 1h, 12h and 1d are always one request (100 max).
• Topic, top-story and location feeds return what Google shows now (35-70 articles) and cannot be filtered by time or site.

Speed and reliability (measured)

Run	Articles	Time	Failed requests
Cloud, default input (1 query, 20 articles)	20	5 s	0
Local, 3 queries x 300 (7 days) + 2 topics + 1 location	1,060	3.4 s	0
Local, 1 query, 30 days, 1,000 articles	1,000	3.9 s	0
Cloud, 1 query x 100 with link decoding	100	18 s	0 (100 of 100 links decoded)
Cloud, 3 queries x 150 + 1 topic with link decoding (earlier build, 10 parallel decodes; now 30)	515	163 s	0 (515 of 515 links decoded)

Filled on every row in these runs: title, sourceName, sourceUrl, publishedAt, url, googleNewsUrl, articleId. Link decoding is deliberately rate-limited (at most about 12 links a second), so plan on roughly 5 articles per second with it on.

API example

Python (pip install apify-client):

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("rel8ble/google-news-scraper").call(run_input={"queries": ["nvidia", "openai"], "timeRange": "1d", "language": "en-US", "country": "US", "maxItemsPerQuery": 100, "resolveUrls": True})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["publishedAt"], "|", item["sourceName"], "|", item["title"], "|", item["url"])

One HTTP call that runs the actor and returns the rows (curl):

curl -X POST "https://api.apify.com/v2/acts/rel8ble~google-news-scraper/run-sync-get-dataset-items?token=<YOUR_APIFY_TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{"queries": ["artificial intelligence"], "timeRange": "1d", "maxItemsPerQuery": 50}'

Integrations

• API: start runs and read datasets over the Apify REST API or the JavaScript / Python clients.
• MCP / AI agents: call it from Claude, ChatGPT or any MCP client through the Apify MCP server; inputs and outputs are documented field by field.
• No-code: Make, Zapier, n8n, Google Sheets, Airtable, Slack, webhooks.
• Scheduling: run hourly or daily in Apify Schedules with time range 1h / 1d for a rolling news feed.

Limits

• No article text, author or image: Google News lists headlines, and its feed carries no excerpt. Turn on link decoding and send publisherUrl to an article extractor for the body text.
• publishedAt is the time Google News reports. For older articles Google often gives only the day (time 07:00 or 08:00 UTC).
• Results depend on the language and country edition; the same query returns different outlets for US and GB.
• An unknown place name in Locations still returns a feed, because Google matches it to the nearest place it knows. Check feedUrl if a location looks off.
• If Google cannot decode a link, the article is still saved with urlResolved: false, the Google link in url and the reason in resolveError.
• Custom Google News topic IDs and "For you" feeds are not supported.

How it works (for developers)

Feeds: https://news.google.com/rss/search?q=<query>&hl=<lang>&gl=<country>&ceid=<country>:<lang> for searches, /rss for top stories, /rss/headlines/section/topic/<NAME> for sections and /rss/headlines/section/geo/<place> for locations. Time and site filters are Google search operators appended to the query: when:7d, after:2026-09-01 before:2026-09-06, site:reuters.com. Each <item> gives title, link, guid, pubDate and source; the <description> is an HTML list of the story cluster, which becomes relatedArticles. Deep runs chain day slices (after:D before:D+1), four in parallel per query, until Max articles per query is reached.

Link decoding: GET /rss/articles/<id> returns a page with data-n-a-sg (signature) and data-n-a-ts (timestamp); a POST to /_/DotsSplashUi/data/batchexecute (rpc Fbv4je) with the ID, timestamp and signature returns the publisher URL.

Run it locally:

npm install
npm test                 # 16 parser tests on saved real feeds and decode responses
node src/main.js         # input in storage/key_value_stores/default/INPUT.json
node scripts/fill.mjs    # field fill rates of the local dataset