Google News Search: headlines, publishers and publication dates

Concept illustration; article images are not downloaded.

Collect structured Google News metadata from keyword searches or category feeds for brand monitoring, industry research and news discovery. Enter a search phrase to receive headlines, publishers, publication dates and Google News article links. No Google account, external API key or proxy configuration is required.

Article events cost $0.35–$0.50 per 1,000 delivered articles depending on your Apify pricing tier, with no start fee. Apify platform usage is charged separately. Begin with five rows: article events cost at most $0.0025 at the listed rates, plus platform usage.

Start with five headlines

{
  "queries": ["OpenAI"],
  "maxResults": 5
}

Run the actor, then open the Dataset. Each row represents one article. Download JSON, CSV or Excel through Apify. Open the SUMMARY key-value record to see which queries were processed and whether a limit was reached; ERRORS contains diagnostic records.

What you can collect

Need	Supported behavior
Brand or company news	Search one or more phrases in the public Google News RSS feed
Industry research	Collect headlines and publisher metadata for topic queries
Regional news	Set a language and country, such as French / France
A publication window	Filter the returned snapshot by UTC publication date
Several search terms	Merge results and deduplicate identical Google News article paths
Export for an analyst or agent	Stable field names, explicit missing values and separate diagnostics

Each RSS feed supplies a bounded snapshot, commonly up to approximately 100 items. Snapshot mode reads one feed per query; daily mode requests separate date-restricted feeds and merges their results. Neither mode is an exhaustive news archive. A local date filter cannot restore items missing from a returned feed. Search ranking, completeness and cross-locale article identity are controlled by the source.

Choose settings for your workflow

Workflow	Start with	Useful fields	Next step in your system
Brand news monitoring	One company name, five results	title, publisherName, publishedAt	Review relevance before sending alerts
Competitive intelligence	Separate brand queries and a per-query cap	query, matchedQueries, articleId	Group coverage by brand and publisher
Industry news research	A focused topic and an appropriate locale	snippet, googleNewsUrl, publishedAt	Open the source before making claims
Regional news discovery	Local-language phrase, language, country	Locale fields and publisher metadata	Check that the source covers your region
AI research assistant	A small JSON input and explicit row budget	All structured fields plus diagnostics	Cite the returned link and verify the article
Spreadsheet reporting	A bounded batch, exported as CSV	Headline, publisher, UTC date and link	Deduplicate imported rows by articleId

This Google News scraper retrieves metadata from public RSS search feeds. It suits news discovery and lightweight monitoring when a current snapshot is sufficient. The actor does not perform sentiment analysis, verify claims or decide whether an article is relevant to your business.

Monitor repeatedly without repeated alerts

Schedule runs in your Apify account or invoke the API from your automation platform. Keep a destination table keyed by articleId. On the first run, save a baseline; on later runs, alert only on IDs your destination has not seen. Store the article's publishedAt separately from collectedAt so an older story appearing in a new snapshot is not mistaken for a newly published story.

This workflow requires state in your own database, spreadsheet or automation. The actor itself does not persist a seen set between runs. Retrieving the same article again produces another delivered result and another article event. Google may assign different article URLs to related coverage, so deduplication is not semantic story clustering.

Map results into a spreadsheet

Use title as the headline column, publisherName as the source, publishedAt as the publication timestamp and googleNewsUrl as the article link. Keep articleId in an additional column for deduplication and query for attribution. Exporting the Dataset does not add a separate actor event; platform storage or transfer charges can still apply.

Do not use publisherUrl for the article-link column: it is a publisher homepage. Preserve null values so downstream users can distinguish a missing value from an empty string or a fabricated URL.

A simple form for the first run

1. Choose Search by company or topic phrase and enter a phrase, or choose Browse Google News categories and select categories.
2. Keep How many articles do you need? at five and run.
3. Optionally open Language, country and freshness to select readable language/country names, a rolling period or newest-first order.
4. Leave Exact dates and collection controls unchanged unless you need a specific UTC window or network limits.

Language and country selectors suggest common editions while accepting custom codes. Existing API calls with language and country remain compatible.

Every input setting

Field	Default	Meaning and limits
queries	Required in search mode	Array of 1–20 search phrases, 1–300 characters each. Exact duplicate phrases are trimmed and removed. Queries are processed in order.
maxResults	5	Maximum unique article rows across the run, 1–1,000. Later queries are skipped once this cap is reached. This cap does not guarantee that many results.
timeRange	"any"	any, 24h, 7d or 30d. Requests a recent feed and applies a rolling UTC lower boundary locally. Intersects exact dates. Not historical pagination.
sortBy	"relevance"	Source order or "newest". Newest sorts each feed before caps and then the selected rows; query order still affects allocation. Unknown dates go last.
language	"en"	Language code, for example en, fr or zh-Hans. A syntactically valid locale may still have limited source coverage.
country	"US"	Uppercase two-letter country code, for example US, FR or GB.
publishedAfter	Unset	Inclusive publication boundary: ISO date or datetime. A date means midnight UTC; a datetime without an offset is interpreted as UTC.
publishedBefore	Unset	Exclusive publication boundary, with the same format. Must follow publishedAfter.
maxArticlesPerQuery	100	Maximum matching candidates examined per query after date filtering, 1–1,000 across all its feeds. Candidates already seen in another query count toward this cap.
timeout	20	HTTP timeout per attempt, 5–45 seconds. This is separate from the total Apify run timeout.
maxRetries	1	Retries per feed for transient network failures, HTTP 429 and selected server errors; 0–2. Backoff is bounded.
mode	"search"	search for phrases or topics for native category feeds.
topics	["TECHNOLOGY"]	1–8 native category codes, used in topics mode. Search phrases are ignored in this mode.
searchMode	"snapshot"	snapshot or daily; daily is available only for keyword search.
daysBack	7	1–31 days back from the exact end or now when no exact start is supplied.
maxFeedRequests	20	Global RSS fetch cap, 1–60, excluding retries and optional article requests.
resolvePublisherUrls	false	Attempt direct publisher article URLs for the selected enrichment rows.
includeFullText	false	Also resolves URLs and attempts public-page main-text extraction.
maxEnrichedArticles	5	Enrich only the first 1–100 selected rows; remaining rows are explicitly skipped.
maxTextCharacters	20000	Saved text cap per article, 1,000–100,000 characters.

Only the query and total limit are needed in default search mode. Other settings are optional.

To give every query a chance, choose a larger global cap or make separate runs.

Useful examples

Newest headlines from the last 24 hours

{
  "queries": ["OpenAI"],
  "maxResults": 5,
  "timeRange": "24h",
  "sortBy": "newest"
}

The period is relative to the run start. It requests recency from Google and checks each publication timestamp locally; it does not promise complete coverage. With multiple queries, early queries can still consume the global cap. For historical dates, keep timeRange at any and use the exact date boundaries.

Compare two company searches

{
  "queries": ["OpenAI", "Anthropic"],
  "maxResults": 10,
  "maxArticlesPerQuery": 5
}

Find French technology headlines

{
  "queries": ["intelligence artificielle"],
  "language": "fr",
  "country": "FR",
  "maxResults": 5
}

Filter a returned snapshot by date

{
  "queries": ["renewable energy"],
  "publishedAfter": "2026-10-01",
  "publishedBefore": "2026-11-01",
  "maxResults": 10
}

Replace these example dates with your desired period. An old period may return no matches because RSS is not a historical archive. For repeat monitoring, schedule runs in Apify and deduplicate articleId in your destination; this actor does not persist a cross-run seen set.

Three-topic research batch

{
  "queries": ["battery recycling", "grid storage", "solar manufacturing"],
  "language": "en",
  "country": "GB",
  "maxResults": 15,
  "maxArticlesPerQuery": 5
}

The per-query cap gives later queries a chance before the global cap is filled. It does not guarantee five new articles per topic: overlap, date filters and source availability can reduce the result count.

Output fields and link provenance

Field	Description
recordType	Always article; diagnostic rows are never added to the Dataset
articleId	SHA-256 of the Google News article URL path; stable across tracking-query changes
googleNewsId	Source RSS GUID when present
title	Headline with an exact trailing publisher suffix removed when identified
rawTitle	Original headline text from the feed
googleNewsUrl	Google News article or RSS redirect URL
publisherName	Publisher named by the feed, or null
publisherUrl	Publisher homepage supplied by the feed, or null
publisherArticleUrl	Direct publisher article URL when resolution succeeds; otherwise null. Distinct from publisher homepage.
publishedAt	Parsed publication time in UTC, or null if unparseable
publishedAtRaw	Original publication date string when available
snippet	Plain text extracted from the RSS description; may repeat a headline or aggregate related links
query	First processed query that produced the article
matchedQueries	Queries in the processed snapshot in which this article was observed; not an exhaustive list of all possible matches
language, country	Requested locale, not independently detected article language/location
collectedAt	Time this run started collection, not the article publication time
linkType, dataSource	google-news-redirect and google-news-rss

publisherUrl is a homepage. googleNewsUrl remains the original redirect even when publisherArticleUrl is resolved. Optional text extraction fetches public publisher HTML and may expose author/image metadata when present; it does not bypass login or paywalls. Different Google News URLs for the same story can remain separate rows.

Precise configuration for AI agents

Use the input keys exactly as documented. queries is an array of strings, not a URL list; maxResults is a global result cap, not a page size. Start with five results, inspect the schema and diagnostics, then increase limits for the actual research task. Use resolvePublisherUrls or includeFullText only when needed and check each row's status. includeFullText also enables link resolution. Optional work is bounded by maxEnrichedArticles.

For a multi-query request, decide whether query order matters. An early query can consume the global cap, leaving later queries unprocessed. Use a small maxArticlesPerQuery, a larger global cap or separate calls when you need a clearer per-topic allocation. matchedQueries records observed matches among processed queries, not all searches that could match an article.

Treat publishedAfter as inclusive and publishedBefore as exclusive. For one UTC calendar day, use the start of that day and the next day as boundaries. The filter is applied to the returned RSS snapshot and cannot turn the source into a historical archive. If a requested date range is empty, report the bounded search outcome rather than claiming no such news exists.

Before consuming a run automatically, inspect its terminal status, SUMMARY and ERRORS. A row is research metadata, not a verified fact or permission to reproduce the article. Use googleNewsUrl for source inspection; do not construct a direct publisher URL from the homepage or headline.

Diagnostics, limits and failures

• SUMMARY.status is SUCCEEDED, PARTIAL or FAILED. It includes per-query source counts, skipped queries, global limits and the number of delivered rows.
• fullSearchCoverage is always false. A successful run proves that a bounded feed was processed, not that every matching news article was collected.
• A valid empty feed or date-filtered empty result is a successful run with zero article rows. HTML, blocked requests, malformed XML and invalid article rows are recorded as errors.
• If one feed fails after another returns useful rows, the useful rows remain available and the run fails with partial-result diagnostics. Check ERRORS before consuming a batch automatically.
• Unknown dates remain null; rows with unknown dates are excluded when a date boundary is active.
• A delivery budget limit is reported separately in SUMMARY.deliveryLimited. Do not use a run status alone to infer complete delivery.

Pricing by Apify tier

One useful article delivered to the default Dataset is one apify-default-dataset-item event. Your Apify pricing tier determines the rate; collecting more rows does not automatically unlock a different tier.

Apify pricing tier	Per article	Per 1,000 delivered articles	Reduction from FREE rate
FREE	$0.00050	$0.50	—
BRONZE	$0.00045	$0.45	10%
SILVER	$0.00040	$0.40	20%
GOLD	$0.00035	$0.35	30%
PLATINUM / DIAMOND	$0.00035	$0.35	30%

These are Apify billing tier identifiers. Check the Pricing tab for the tier applied to your account. Enterprise tiers currently use the same article rate as GOLD.

There is no start fee and no extra actor event for queries, deduplication, locale selection or date filtering. Duplicates removed within a run, diagnostics and empty results create no article events. Articles collected again in a later run are new delivered results and are charged again.

Delivered articles	FREE article events	GOLD article events	Additional cost
0	$0	$0	Platform usage for the attempted run
5	$0.00250	$0.00175	Platform usage
100	$0.050	$0.035	Platform usage
1,000	$0.50	$0.35	Platform usage
10,000 across runs	$5.00	$3.50	Platform usage across runs

The examples are arithmetic, not promises of source coverage. Each feed provides a bounded snapshot; daily mode can merge several feeds. A single run returns at most 1,000 rows.

Apify compute, storage, transfer and applicable platform services are paid by the user at the user's plan rates. No residential proxy or external service key is required. See each run's usage for the total. Set both a maximum total charge and a sensible run timeout; a result charge cap is not a universal cap on platform usage. When a delivery budget stops output, inspect SUMMARY.deliveryLimited before using the results.

Browse native Google News categories

Select Browse Google News categories, then choose World, Nation, Business, Technology, Entertainment, Sports, Science or Health. These requests use Google News RSS topic sections rather than keyword searches named after categories. Country/language choose the edition; Nation is edition-specific. Query strings are ignored in this mode.

{
  "mode": "topics",
  "topics": ["TECHNOLOGY", "SCIENCE"],
  "maxResults": 10,
  "maxArticlesPerQuery": 5
}

Rows carry topic and matchedTopics; query is null and matchedQueries is empty for a category result. Overlapping category URLs are deduplicated within the run. Limits and rolling/exact publication filters still apply. Category feeds cannot be swept day-by-day by this version.

Search separate days

Daily mode sends distinct date-restricted RSS requests, newest day first. It applies each window's UTC boundaries locally, combines useful results and removes duplicate article IDs. It can discover articles absent from one current snapshot; it does not expose a next-page cursor or prove full historical coverage.

{
  "queries": ["OpenAI"],
  "searchMode": "daily",
  "daysBack": 3,
  "maxFeedRequests": 4,
  "maxArticlesPerQuery": 250,
  "maxResults": 50
}

With an exact start/end, daily mode uses that interval instead of the automatic start. The interval is limited to 31 elapsed days, which can span 32 partial calendar days. timeRange intersects the exact dates; choose any for a historical interval. An explicit start more than 31 days before the end is rejected before collection.

maxFeedRequests caps the entire run's RSS fetches; retries do not count as new windows, but use network resources. Inspect SUMMARY.feedRequests, requestLimitReached and each feed's sourceUrl, windowStart, windowEnd and status. A global row cap or per-query candidate cap can stop collection before all windows or later queries are attempted. Source omissions remain possible even when every planned window is attempted.

Direct publisher links and article text

{
  "queries": ["site:science.nasa.gov Webb"],
  "maxResults": 3,
  "includeFullText": true,
  "maxEnrichedArticles": 3,
  "maxTextCharacters": 20000
}

For links only, set resolvePublisherUrls: true and leave includeFullText off. For text, one checkbox enables both resolution and public-page reading. The actor uses Google News decoding requests, then fetches publisher HTML for the selected rows. No external API key or residential proxy configuration is required. Google rate limits and publisher blocks can still prevent success.

publisherUrlStatus is not_requested, resolved, unresolved or skipped_limit. An unavailable direct URL stays null; the actor does not substitute the publisher homepage. linkType continues to describe googleNewsUrl, so it remains google-news-redirect.

fullTextStatus is not_requested, extracted, url_unresolved, skipped_limit, fetch_failed, insufficient_text, blocked_page or paywall_detected. Extracted text is the main readable content found in the fetched public HTML, not a guarantee of the complete original article. Detected structured paywalls are skipped; login, anti-bot pages, non-HTML responses and insufficient text are not returned as successful article bodies.

Additional output	Meaning
topic, matchedTopics	Native topic provenance, or null / empty array for keyword rows
publisherUrlStatus	Whether direct article resolution was attempted and succeeded
fullText, fullTextStatus	Extracted main text or null, plus an explicit status
fullTextTruncated	Whether extracted text was clipped at your saved-character cap
extractedTextCharacters	Saved character count; zero when no text was extracted
articleFetchUrl	Final publisher-page URL used for successful extraction
author, articleImageUrl	Optional publisher metadata; null when unavailable. Images are not downloaded.

SUMMARY.publisherUrlsResolved, fullTextsExtracted, enrichmentWarnings, enrichmentLimited and enrichmentComplete describe optional-work outcomes. Useful metadata remains delivered if optional enrichment fails; the run can succeed with unresolved links or missing bodies. Check these diagnostics instead of equating a successful metadata run with successful text extraction for every row.

There is no additional actor event for resolution, daily searches or text: one delivered useful metadata row remains one article event. These modes make more network requests, take longer and increase platform usage paid by the user. Start with 1–3 enriched articles and a suitable total runtime. A zero-result event bill can still have platform costs. Large enrichment batches may exceed a short overall run timeout before output is delivered.

API, Python and workflow automation

Use the exact identifier apivault_labs/google-news-search, or Actor ID IWr1Q1aTT3WUatVOl. The JSON inputs above are also the contract for automation agents. Use an Apify API token from your own account and keep it in an environment variable or workflow credential.

Python with the official Apify client

Install apify-client version 3.3 or later and set APIFY_TOKEN in your environment:

import os
from datetime import timedelta
from decimal import Decimal
from apify_client import ApifyClient

client = ApifyClient(os.environ["APIFY_TOKEN"])
run = client.actor("apivault_labs/google-news-search").call(
    run_input={"queries": ["OpenAI"], "maxResults": 5},
    max_total_charge_usd=Decimal("0.02"),
    run_timeout=timedelta(seconds=120),
)
if run is None or run.status != "SUCCEEDED":
    raise RuntimeError("Run incomplete; inspect its SUMMARY and ERRORS records")
store = client.key_value_store(run.default_key_value_store_id)
summary_record = store.get_record("SUMMARY")
errors_record = store.get_record("ERRORS")
if summary_record is None or errors_record is None:
    raise RuntimeError("Missing diagnostics")
summary, errors = summary_record["value"], errors_record["value"]
if errors or summary["status"] != "SUCCEEDED":
    raise RuntimeError("Incomplete delivery; inspect SUMMARY and ERRORS")
for article in client.dataset(run.default_dataset_id).iterate_items():
    print(article["title"], article["googleNewsUrl"])

A successful run can still hit a global or delivery limit. Read SUMMARY and ERRORS from the run's default key-value store before treating the result as complete. See official Python client documentation and the Actor's API tab for other languages and exact endpoints.

HTTP, n8n, Make and custom agents

1. Send a JSON POST to https://api.apify.com/v2/acts/apivault_labs~google-news-search/runs?memory=256&timeout=120&maxTotalChargeUsd=0.02 with header Authorization: Bearer YOUR_APIFY_TOKEN and Content-Type: application/json.
2. Save the returned run ID. Poll GET https://api.apify.com/v2/actor-runs/RUN_ID until a terminal state, or use an Apify run-finished webhook.
3. Read SUMMARY and ERRORS through https://api.apify.com/v2/key-value-stores/STORE_ID/records/SUMMARY and the equivalent /ERRORS endpoint. Read article rows from https://api.apify.com/v2/datasets/DATASET_ID/items?clean=true.

In n8n, these are standard HTTP Request nodes with stored header credentials. See Apify API documentation. A purpose-built verified n8n node, custom PyPI package and dedicated MCP server are not provided with this version. For MCP clients that expose Apify actors, pass this exact Actor ID and schema, then inspect the diagnostics just as an HTTP client does. No public preconfigured Tasks have been created in this pass.

FAQ

Can I get direct publisher URLs or full text? Enable resolvePublisherUrls or includeFullText. Both are best effort and capped independently of metadata rows. Check publisherUrlStatus, fullTextStatus and SUMMARY.enrichmentWarnings; unavailable fields stay null.

Why did a later query return no rows? Inspect SUMMARY.feeds. The global cap may have skipped the query; its results may also be duplicates or excluded by your date window.

Can this search all articles published last year? No. Daily mode accepts a bounded range of up to 31 elapsed days and may return historical results Google still exposes, but does not guarantee archival completeness. Source/request/result limits still apply.

Can an AI agent choose precise settings? Yes: use the complete parameter table, keep maxResults small first, and read both SUMMARY and ERRORS. Missing data is null; a source failure is not a successful empty result.

Does the country filter prove where an event happened? No. language and country select the requested Google News edition. They are not detected article attributes or a geolocation classifier.

Why is my snippet short or similar to the headline? RSS descriptions vary by source and can contain related links. The actor extracts plain text from that description; it does not download the publisher's full article.

Can I collect 10,000 stories in one run? The input cap is 1,000 rows per run and the source is bounded per query. Multiple runs may overlap and do not establish exhaustive coverage. Price examples for 10,000 rows describe arithmetic across runs.

Does a higher Apify tier change the data? The listed tier changes the article-event price. The actor's search inputs, output fields and source coverage are the same.

What should I check when fewer rows arrive than requested? First inspect skipped queries, per-query limits, date filters, duplicates, source errors and deliveryLimited in the diagnostics. Increasing a cap cannot create results absent from the returned feed.

How should I choose a query? Begin with a specific company or topic phrase. Inspect five results for relevance, then refine the phrase or locale. Search interpretation and ranking come from Google; this actor does not guarantee a particular rank or keyword matching rule.