Google News Scraper
Pricing
from $2.50 / 1,000 articles
Google News Scraper
Collect Google News headlines by search, topic, edition or location. Return publisher, publication time, story clusters and matched searches, with optional publisher links, article text and scheduled new-only delivery.
Pricing
from $2.50 / 1,000 articles
Rating
0.0
(0)
Developer
Philip Kirkbride
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
3 days ago
Last modified
Categories
Features
Share
Collect Google News headlines for sports coverage, company monitoring, research and news alerts. Search with Google's news operators, choose a topic or location, and select a language/country edition. Each article includes its publisher, publication time, Google News link and the searches it matched. Publisher article links and readable article text are optional.
Use cases
- Build a sports news writing queue from team, league or player searches.
- Send trading research systems timestamped company and market headlines. Verify a story before acting on it.
- Monitor competitors and selected publishers with domain filters.
- Schedule a new-article feed for daily newsletters and news alerts.
Also available
The Google data suite: pick the job.
| Actor | Use when |
|---|---|
| Google News (this Actor) | Headlines by topic, search or edition |
| Google Search Results | Rankings, snippets and AI Overviews per query |
| Google Trends | Search interest over time and by region |
| Wikipedia Pageview Spikes | Articles whose daily views spiked |
| Posts and comments from subreddits and searches |
Quick start
Run the Console prefill for 10 US English top headlines:
{"topHeadlines": true,"language": "en-US","country": "US","maxItemsPerQuery": 10,"maxResults": 10,"resolvePublisherUrls": false,"fetchArticleText": false,"onlyNewSinceLastRun": false,"deduplicate": true,"proxyConfiguration": {"useApifyProxy": false}}
A bare input {} also delivers a sample of up to 10 current top headlines. Its status begins Sample run (...) and OUTPUT.sampleRun is true. Language, country and other option settings alone still use this sample; select a source or set a result cap to start an explicit run. An explicit cap without sources selects top headlines and honors that cap. An explicit empty list, an empty string or a null field ends the run with no rows and names the field to fix. A single entry a source cannot use (a blank query, an unknown topic, a malformed domain) is skipped and listed in OUTPUT.skippedSearches while the rest still run; a source list where every entry is unusable ends the run with no rows and says so.
For two small searches from the last day, run:
{"queries": ["football","basketball"],"publishedWithin": "1d","maxItemsPerQuery": 3,"maxResults": 6}
These examples were run locally on November 11, 2026. Search results change over time.
Inputs
All inputs are optional. Supply one or more of queries, topics, topHeadlines: true or locations to select sources. Sources combine in one run. Unknown fields and incorrect types fail before collection.
| Field | Omitted behavior | Description |
|---|---|---|
queries | No searches | 1–20 nonempty strings, each up to 500 characters. Quotes, OR, -term, site:, intitle:, when:, after: and before: pass through. |
topics | No topics | 1–20 choices from WORLD, NATION, BUSINESS, TECHNOLOGY, ENTERTAINMENT, SPORTS, SCIENCE, HEALTH. |
topHeadlines | False; true for sample | Include current top headlines for the selected edition. |
locations | No locations | 1–20 place names, each up to 500 characters. Google decides geographic coverage. |
language | en-US | Exact supported edition language code; see the edition table. |
country | US | Exact uppercase supported edition country code. |
publishedWithin | No recent-window filter | 1h–99h or 1d–31d, for example 6h. Filters by publication time in UTC. |
dateFrom | No date range | Inclusive UTC date YYYY-MM-DD; requires dateTo and search queries only. |
dateTo | No date range | Inclusive UTC date YYYY-MM-DD; requires dateFrom. At most 31 inclusive days. |
sites | All publishers | 1–20 lowercase domains, without schemes or paths. Includes their subdomains. |
excludeSites | No exclusions | 1–20 lowercase domains to exclude, including subdomains. Cannot overlap sites. |
maxItemsPerQuery | 100; 10 for sample | 1–1,000 articles considered per original source across all date slices, before cross-source deduplication and saved-state filtering. |
maxResults | 100; 10 for sample | 1–5,000 total delivered rows. Sample is always capped at 10. |
resolvePublisherUrls | False | Recover publisher article links where possible; failures keep the Google News link and include a reason. |
fetchArticleText | False | Fetch readable publisher text and metadata. Automatically enables URL resolution. |
onlyNewSinceLastRun | False | Deliver only article IDs not previously delivered under the same state key. |
stateKey | Derived from sources, edition and domain filters | Nonempty name of up to 64 characters for a scheduled feed. Use a new name for a fresh baseline. |
deduplicate | True | Deliver an article once across sources. matchedQueries lists matching source labels known when that row was delivered. Matches found by later feeds are retained in OUTPUT.additionalMatches, keyed by article ID. |
proxyConfiguration | Direct connection | Optional connection settings on Apify. Each custom proxy/group list permits 1–20 nonempty strings, up to 2,048 characters each. Local runs permit only the disabled setting shown in the prefill. |
Duplicate list entries are trimmed and collapsed in first-occurrence order. Non-Latin search text is supported; whitespace-only strings fail. Lists and result caps use Console prefill values rather than injected defaults. An explicit run without caps can return up to 100 rows; the prefill requests only 10.
Supported editions
Use one of these exact language/country pairs. Changing country does not automatically change language.
| Language | Countries |
|---|---|
en-US | US |
en-GB | GB |
en-CA | CA |
en-AU | AU |
en-IN | IN |
es | ES, PE |
es-419 | MX |
pt-BR | BR |
fr | FR |
de | DE |
it | IT |
ar | SA, EG |
he | IL |
ja | JP |
ko | KR |
zh-CN | CN |
zh-TW | TW |
hi | IN |
All 20 listed language/country pairs were checked against live feeds on November 11, 2026. The edition affects Google's selection and language preference; it does not translate articles or guarantee every result uses that language.
Time windows and coverage
For searches, publishedWithin adds Google's when: operator and also filters returned publication timestamps locally. For topics, top headlines and locations it filters the current feed locally. It cannot recover articles that have already left that feed.
A dateFrom/dateTo search automatically splits into one request per UTC day. Each search uses broadened after:/before: boundaries, then keeps only that slice's exact UTC day. OUTPUT.dateRangeSplitByDay and dateSlices show the split. If you also set publishedWithin, both restrictions apply. Operators already inside a query remain in place and can further narrow the results.
Each Google News RSS feed returns at most about 100 items; 104 were observed in one live search. There is no next-page collection. Daily splitting helps with wide windows but a busy day can still omit articles. Split busy searches further by publisher or keyword. Historical dates are supported for searches only; dated topic, headline and location requests fail with a message naming the date fields. A run can plan at most 60 feeds. Combined queries/domain filters must fit an 8,000-character encoded source URL; oversized combinations fail before collection.
Domain filters add site: restrictions to searches and also check each returned publisher domain locally. For other feed types they filter the current feed locally. An article without a known publisher domain cannot pass an include-domain filter.
Output example
One complete headline record captured from the two-search example above:
{"articleId": "b1ca1abf17d836fdab8205cf44deabbdef8bc6611c472a8a01bebc265c1dbcc5","title": "College Football Playoff bracket, based on current committee rankings","publisher": "NCAA.com","publisherUrl": "https://www.ncaa.com","googleNewsUrl": "https://news.google.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?oc=5","resolvedUrl": null,"resolutionStatus": "not_requested","resolutionReason": null,"publishedAt": "2026-11-11T21:45:51+00:00","description": "College Football Playoff bracket, based on current committee rankings NCAA.com","relatedArticles": [],"query": "football","matchedQueries": ["football"],"feedType": "search","topic": null,"language": "en-US","country": "US","collectedAt": "2026-11-11T22:08:56.962111+00:00","article": null}
The default dataset contains one row per delivered article. Cross-source duplicates share one row when deduplicate is true; the first match supplies query, feedType and topic. With deduplication off, an article can have multiple rows, each retaining every matched source label. New-only mode always delivers each ID once per run.
| Field | Meaning |
|---|---|
articleId | Stable SHA-256 of Google's item ID, or its link ID when no item ID is supplied. Different Google IDs for the same publisher story remain different articles. |
title, publisher, publisherUrl | Article headline, publisher name and publisher home page, as provided by Google. Publisher fields may be null. |
googleNewsUrl | Original Google News article link, always retained. |
resolvedUrl | Recovered publisher article URL; null when not requested or unavailable. |
resolutionStatus, resolutionReason | Status: not_requested, offline, redirect, lookup or failed; failure reason when applicable. |
publishedAt | Google's item publication timestamp, ISO 8601 in UTC. Items with unusable publication times are counted and skipped. |
description | Google's description with HTML removed. Often a headline or a list of related headlines, rather than a prose summary. |
relatedArticles | Other linked stories in Google's item description, each with title, publisher and link. Not separately delivered or fetched. |
query, matchedQueries | First matching source label and all matching labels observed in the fetched feeds. Non-search labels are the topic, location or “Top headlines”. |
feedType, topic | search, topic, topHeadlines or location; topic code only for a topic row. |
language, country | Requested edition. |
collectedAt | Collection timestamp, ISO 8601 UTC. |
article | Null unless text was requested. Includes title, text, author, publishedAt, image, canonicalUrl, wordCount and fetchStatus. |
Article extraction uses readable content available on the publisher page. Metadata can be null; publisher publication times without a timezone are left null. wordCount counts whitespace-separated tokens and can undercount CJK text. A successful extraction does not certify that a publisher exposed its entire article. Restricted subscriptions, access refusals, non-HTML responses, short or missing article content, oversized pages and timeouts have distinct failure statuses and null text. The headline row is still delivered. Text collection has a 2 MB response limit, a 20-second HTTP timeout and a per-domain delay. A mixed-publisher check on November 11 extracted text from only 3 of 6 pages; the other pages reported a paywall or missing text. Success varies substantially by publisher and network, so do not assume that most pages will yield text. URL and text enrichment process at most six articles concurrently, with request pacing and shared backoff after rate limits.
The OUTPUT key-value record is an uncharged run summary: delivered and attempted rows, per-source/date counts, HTTP request and response-byte counts, failure reasons, enrichment successes, sample flag, cap indicators, saved-state information and summaryLine. Source-feed cap indicators are approximate. Per-feed delivered counts include shared matches and can sum above dataset rows. A terminal status repeats the summary. If every feed fails, the run fails with the reasons; partial failures remain visible alongside delivered rows. OUTPUT.status is succeeded, succeeded_partial (some inputs were skipped), no_results (the feeds were read but matched no articles) or failed.
Scheduling and new-only feeds
Schedule the same input on Apify and enable onlyNewSinceLastRun. This live-tested small baseline is suitable for trying the behavior:
{"queries": ["site:theguardian.com football"],"maxItemsPerQuery": 3,"maxResults": 3,"onlyNewSinceLastRun": true,"stateKey": "live-gnews-feed"}
The first run delivers its baseline. The second identical local run delivered zero and charged no article events. Adding a second search under the same state key delivered three previously unseen IDs. Timing and changes in Google selection can change these counts.
The named store remembers up to 20,000 IDs for 90 days after delivery, independently of a run's default storage. Expired or evicted IDs can be delivered again. OUTPUT reports feed.expiredIds and feed.evictedIds; an eviction also warns in the terminal status. The most recently delivered IDs win retention ties. Only stored rows enter the ledger, including when a spending limit stops a batch. Reading corrupt state fails rather than silently resetting it. A state-save failure is reported and may cause repeat deliveries next time. A concurrent run with the same state key fails before source collection with a stateKey busy message. Different state keys can run concurrently. After an interrupted run, its exclusive lease can take up to 15 minutes to expire. The lock prevents overlapping writers; it cannot make dataset delivery and state persistence one atomic transaction. A forced termination or state-save failure between them can cause repeat deliveries, so consumers should also deduplicate by articleId. Older or unsupported state shapes fail with a field-named message; use a new state key rather than silently resetting history.
The per-source article cap applies before saved-state filtering. Set it high enough to cover each current feed. If total eligible articles exceed maxResults, the next run can deliver additional previously undelivered IDs from the same source window. The ledger tracks IDs, not headline edits. Use a new stateKey when you want to resend a baseline. Export or process dataset rows through the usual Apify dataset integrations; OUTPUT is a summary, not a dataset row.
Pricing
$0.0025 per article ($2.50 per 1,000) plus $0.005 per run start; resolving publisher URLs and fetching article text are included at no extra charge
Under pay per event there is no separate compute or proxy charge. The start fee applies even when no articles are delivered. Apify counts starts per GB of memory; the default 512 MB run is one start event. Memory is capped at 1,024 MB to keep that count at one.
A 10-article sample costs $0.030 ($0.025 in article events + $0.005 start). A 100-article run costs $0.255; 1,000 delivered articles in one run cost $2.505. A zero-new scheduled check costs $0.005. Splitting the same articles across multiple runs adds a start fee each time.
One dataset item is delivered per article. Cross-source duplicates are suppressed when deduplicate is true, and new-only mode suppresses previously delivered IDs. The delivered prefix stops at the run's spending allowance and the status states “allowed N of M”. A spending allowance of $0.010 permits the start and two articles; no uncharged article rows are written.
Limits and freshness
Google News RSS is a changing selection of indexed publisher stories, not a complete news archive. Some publishers and articles are absent, clusters change, and publication times can differ from a publisher's page. Google controls freshness and search semantics; there is no real-time delivery guarantee. A zero-result explicit search is valid. Check the summary and publication timestamps before using headlines in time-sensitive systems.
Rows are saved after each feed, before waiting for later sources. Later feeds can add provenance through OUTPUT.additionalMatches; earlier dataset rows are not rewritten. Planned feeds are still inspected for matches after the result cap, unless a charge, time or storage limit stops the run. Optional text and link work runs only for selected new rows. A run keeps at most an eight-minute collection/enrichment budget and reserves 45 seconds before the platform deadline for delivery and a summary. If enrichment runs out of time, collected metadata is still delivered and unfinished enrichment reports time_budget. OUTPUT is checkpointed during collection. A hard stop or abort can still interrupt the final in-flight write. Use at least a 600-second timeout for larger runs and smaller result limits when enrichment is enabled. Daily schedules can miss stories that enter and leave the current feeds between runs.
FAQ
Can I collect more than 100 articles? Combine sources or use an explicit date range, which splits searches by day. Increase both relevant result caps. A single feed still contains only about 100 selected items and cannot be paged.
Does the description contain the article? Usually it contains a headline or related-story list. Enable fetchArticleText for optional publisher text and inspect article.fetchStatus.
Will every publisher URL resolve? No. Failed resolution leaves resolvedUrl: null, keeps googleNewsUrl, and records resolutionReason. No publisher is substituted.
Does a schedule resend yesterday's news? Enable new-only mode and keep the same state key. It suppresses delivered IDs while they remain in the ledger; new Google IDs and expired state can appear again.
Can I export JSON or CSV? Yes. Use the Apify dataset export options; JSON retains nested related-story and article fields most naturally.