google-news-scraper-api

Node client that turns a query into Google News articles as JSON. Built on ScrapingBee's Google News scraper API. Zero dependencies, Node 18+.

Checked against the live endpoint on 2026-08-25. Calls charged 10 credits.

npm install google-news-scraper-api

Build a media monitor

The whole job is four steps: query, page, dedupe, store. Everything below is those four.

1. Query

const { GoogleNewsScraper } = require("google-news-scraper-api");

const scraper = new GoogleNewsScraper(process.env.SCRAPINGBEE_API_KEY);
const articles = await scraper.search("openai", { country_code: "us" });

articles.forEach(a => console.log(a.date, a.source, a.title));

Each article arrives with title, link, snippet, source, domain, date, relative_date, position and page. date is an ISO timestamp, which is the one you want for sorting; relative_date is the "3 hours ago" string meant for humans.

2. Page

const wide = await scraper.paginate("climate policy", 5, { country_code: "us" });

Paging stops the moment a page returns empty, so asking for five pages of a thin query does not spend five pages of credits.

3. Dedupe

Wire stories get syndicated. The same piece shows up under different publishers and across queries, and a monitor that skips this reports the same story five times:

const raw = [];
for (const q of ["openai", "anthropic", "mistral ai"]) {
  raw.push(...await scraper.paginate(q, 2, { country_code: "us" }));
}
const unique = GoogleNewsScraper.deduplicate(raw);
console.log(raw.length, "->", unique.length);

4. Store

searchRaw hands back meta_data alongside the results when you want the resolved Google URL and page count recorded with the batch:

const payload = await scraper.searchRaw("openai", { country_code: "us" });
console.log(payload.meta_data.url);

Two parameters that change the results

await scraper.search("acme corp", { nfpr: true });          // stop Google "correcting" the brand
await scraper.search("acme corp", { date_range: "d" });     // tighten the recency window

nfpr deserves attention on brand monitoring. Google silently rewrites unfamiliar names to something it thinks you meant, and the run then tracks the wrong word entirely.

search_type=news will not run with device: "mobile"; that combination is unsupported.

Budget

| Call | Credits | | --- | --- | | One query | 10 observed, 15 documented standard | | usage() | 0 | | HTTP 500 | 0 |

console.log(await scraper.usage());

Free, six calls a minute. Pricing tiers.

Sibling surfaces

search_type swaps the same endpoint onto search, shopping, images, lens, ads and AI Mode. There is also a Google News RSS API and the lighter Fast Search.

Notes

Public content only; scraping behind a login is prohibited by the ScrapingBee terms. Keep your key away from AI coding assistants.

MIT. Docs . Repo