Google News RSS Feeds: How They Work, Their Limits and Alternatives
Michael Carter · · 14 min read

Key takeaways
- Google publishes no documentation for Google News RSS feeds: the URL formats and the hl, gl and ceid parameters in this guide are what we observed on November 18, 2026, and every feed's copyright notice limits it to a personal feed reader.
- A Google News RSS search feed returned 92 to 104 items per request in our tests, top stories 34 to 38 and a topic or place feed 66 to 70, all with publishers and publish times but no snippets.
- Every item link was a news.google.com article link with an opaque ID: in 2 of 2 tries, plain HTTP got a 302 and then a Google News page that did not contain the publisher's URL.
- For the same four queries, 54 of the 57 Google News tab results that BestSERP returned were also in the RSS feed, with identical publish times; the News tab added final publisher URLs and, on 25 of 57 results, a snippet.
- BestSERP's /news endpoint uses the Google News RSS feed only when Google blocks the News tab, and marks those items with origin google_news_rss.
Google News RSS feeds are free XML feeds of Google News: top stories, topics, places and any search, for a chosen country and language. Google does not document them. In our tests on November 18, 2026, a search feed returned 92 to 104 items with headlines, publishers and publish times, Google redirect links and no snippets.
It is written for developers who want news in a script, a feed reader or an AI pipeline: tested URL formats, what an item contains, measured limits, a Python reader and a side-by-side test against the Google News tab as JSON. It is part of our guide to shopping, local and news data.
Google News RSS in one paragraph
Google News RSS is Google News as RSS 2.0 files under news.google.com/rss: one XML document per feed, up to about 100 items, each with a headline, a link, a publish time and a publisher. No key or account is needed: plain HTTP answered all our requests, with a browser User-Agent and with one that simply names the script. Two facts frame the rest:
- No documentation. We found no Google help or developer page about these feeds or their parameters (checked November 18, 2026). Everything below is observed behavior that Google can change without notice, as it did in 2017, when a help forum post announced that "the old RSS URL patterns will no longer work effective December 1st, 2017" (Search Engine Roundtable, November 3, 2017).
- The terms are in the feed. Every feed we fetched carries the same
<copyright>element: "This XML feed is made available solely for the purpose of rendering Google News results within a personal feed reader for personal, non-commercial use. Any other use of the feed is expressly prohibited." Read it before you build a product on these feeds (general information, not legal advice).
Google News RSS URL formats
A Google News RSS feed URL is the root https://news.google.com/rss, a path for the feed type, and three edition parameters. These are the forms we requested on November 18, 2026:

| Feed | Path (plus ?hl=en-US&gl=US&ceid=US:en) |
Answer |
|---|---|---|
| Top stories | /rss |
200, 38 items |
| Search | /rss/search?q=NASA |
200, 92 to 104 items for our 4 test queries |
| Topic by name | /rss/headlines/section/topic/TECHNOLOGY |
302 to /rss/topics/CAAq..., then 66 items |
| Place by name | /rss/headlines/section/geo/Chicago |
302 to /rss/topics/CAAq..., then 70 items |
Named topics and places are redirects: with redirects turned off, your client gets an empty 302 instead of a feed. The query is URL-encoded (q=electric+vehicles, quotes as %22). Two search operators worked in our tests:
when:1d:Nvidia when:1dreturned 100 items, all published in the previous 15.5 hours. Without it, the oldest of 104 items was 11 days old.site::Nvidia site:cnbc.comreturned 100 items, all from CNBC; 99 were from 2026 and one, a stock quote page, from July 2017.
Third-party guides list more operators, such as after:, before: and intitle:. We did not test them.
Country and language: hl, gl and ceid
A Google News feed by country takes three parameters: hl (language), gl (country) and ceid (the pair, written country:language). The three editions we fetched had almost nothing in common:
| Edition | Parameters | Top stories | Most frequent publisher |
|---|---|---|---|
| United States, English | hl=en-US&gl=US&ceid=US:en |
38 | The New York Times (5) |
| United Kingdom, English | hl=en-GB&gl=GB&ceid=GB:en |
38 | BBC (12) |
| Germany, German | hl=de&gl=DE&ceid=DE:de |
34 | Tagesspiegel, DIE ZEIT, WELT (3 each) |
The US and UK feeds shared 2 of 38 stories; the German feed shared none. Each feed's <language> matched hl.
Missing parameters were filled in by redirect: hl=en-GB&gl=GB answered 302 to the same URL plus ceid=GB:en, and /rss alone answered 302 to the US English edition (we sent Accept-Language: en-US, so we cannot say what chose it). Send all three. For other editions, copy the values from that edition's Google News address. BestSERP's parameters are written differently (gl=us, hl=en); see how the gl and hl parameters work.
What a Google News RSS feed contains
Every one of the 844 items in the 11 feeds we fetched had the same six elements, shown here on a real item:

| Element | What we saw (844 items) | How to use it |
|---|---|---|
title |
Headline plus " - Publisher" on all 844 | Strip the suffix using source |
link |
news.google.com/rss/articles/<ID>?oc=5, 142 to 672 characters |
Not the article URL (see Limits) |
guid |
The same ID, isPermaLink="false"; identical for the same article in different feeds |
A key for de-duplication |
pubDate |
RFC 822 date in GMT, to the second | Parse it as a UTC time |
description |
HTML: a link plus the publisher, or a numbered list of up to 9 articles for a story cluster (160 items) | No summary text in any item |
source |
Publisher name; url is its home page on all 844, with "www." on 630 |
Publisher identity, not the article |
Items are not sorted by date: only about half of consecutive pairs in the search feeds were newest first. Median ages were 7.7 to 17 hours for searches (the oldest item dated from May 2026) and 4.1 to 8.7 hours for top stories. Each feed arrived in 0.8 to 1.5 seconds.
Read a feed in Python
This reader uses only the standard library: it builds the search URL, keeps the raw XML and turns each item into a dictionary.
import sys
import urllib.parse
import urllib.request
import xml.etree.ElementTree as ET
from email.utils import parsedate_to_datetime
from pathlib import Path
UA = "personal-news-reader/1.0 (Python urllib)" # say what you are; no need to pose as a browser
def feed_url(query, hl="en-US", gl="US", ceid="US:en"):
"""Send hl, gl and ceid together: without ceid, Google answered with a 302 that adds it."""
params = urllib.parse.urlencode({"q": query, "hl": hl, "gl": gl, "ceid": ceid}, safe=":")
return f"https://news.google.com/rss/search?{params}"
def fetch(url):
req = urllib.request.Request(url, headers={"User-Agent": UA})
with urllib.request.urlopen(req, timeout=30) as resp: # follows redirects (named topics and places)
return resp.read()
def parse(xml):
items = []
for item in ET.fromstring(xml).find("channel").findall("item"):
source = item.find("source")
title = item.findtext("title")
if title.endswith(" - " + source.text):
title = title[: -len(source.text) - 3]
items.append({
"title": title,
"source": source.text,
"source_home": source.get("url"), # the publisher's home page, not the article
"published": parsedate_to_datetime(item.findtext("pubDate")).isoformat(),
"link": item.findtext("link"), # a news.google.com/rss/articles/... link
"guid": item.findtext("guid"), # the article ID: a stable key for de-duplication
"cluster": item.findtext("description").count("<li>"), # articles in a story cluster, 0 if none
})
return items
if __name__ == "__main__":
sys.stdout.reconfigure(encoding="utf-8") # headlines are Unicode; a Windows console defaults to cp1252
query = sys.argv[1] if len(sys.argv) > 1 else "NASA"
xml = fetch(feed_url(query, *sys.argv[2:5]))
Path("feed.xml").write_bytes(xml) # keep the raw feed
items = parse(xml)
print(f"{len(items)} items, {sum(1 for i in items if i['cluster'])} story clusters")
for n, item in enumerate(items[:5], 1):
extra = f" (cluster of {item['cluster']})" if item["cluster"] else ""
print(n, item["published"], item["source"], "|", item["title"] + extra)
Our run for "NASA" at 07:21 UTC, with the User-Agent above, printed:
92 items, 1 story clusters
1 2026-11-17T11:00:00+00:00 Los Angeles Times | NASA’s Voyager 1 will soon be so far away that its messages take a day to reach Earth (cluster of 9)
2 2026-11-17T13:33:56+00:00 EarthSky | Mars Sample Return: NASA’s new effort to bring samples home
3 2026-11-18T05:42:25+00:00 NASA Science (.gov) | Shine Bright for Voyager’s Light‑Day
4 2026-11-17T17:20:14+00:00 Fox News | NASA investigates mysterious 'potato chip' shapes on Mars as scientists hunt for answers
5 2026-11-17T17:11:52+00:00 NASA (.gov) | NASA Invites Media, Public to View Roman Telescope’s First Images
Poll a feed by comparing every guid with what you stored, not by reading the top items. Fetched again 24 minutes after our first request, the NASA feed had kept 88 of its 92 items and 9 of its top 10; the 4 new items entered at positions 51, 52, 54 and 72, and only 2 of them had been published in between. One kept item had a new pubDate, 48 minutes later than before.
Limits of Google News RSS
The limits of Google News RSS are mostly what it leaves out:
- Links never reached the publisher. Each article ID decodes to an opaque token, not a URL (844 of 844). With plain HTTP, a link answered 302, then a 609 KB Google News page without the publisher's URL (2 of 2 tries); a browser gets there by running the page's JavaScript.
- No snippets. The description holds headlines and publishers only: no summary to show or to give a model.
- About 100 items. Our search feeds held 92 to 104; no URL form we tried returned more.
- Relevance order, with old items. Without
when:, a feed mixes the last hour with articles months or years old. - One edition per request. Each country and language is a separate feed.
- No contract. No documentation, no stated rate limit (our 21 requests, at least 10 seconds apart, all answered), and terms that restrict the feed to a personal feed reader.
Google News RSS vs the News tab: a side-by-side test
We sent four queries to both on November 18, 2026, 06:50 to 06:53 UTC, within two seconds of each other: the US English RSS search feed, and BestSERP's /news (the Google News tab as JSON) with gl: "us", hl: "en" and noCache: true. All four /news answers came from the News tab (origin: "google_search") and from a US IP (ipCountry: "us").
| Query | RSS items (1 request) | /news items (1 credit) | /news items also in the feed | /news items with a snippet |
|---|---|---|---|---|
| Nvidia | 104 | 18 | 16 | 7 |
| electric vehicles | 102 | 13 | 13 | 9 |
| NASA | 92 | 10 | 10 | 1 |
| Federal Reserve | 100 | 16 | 15 | 8 |
| Total | 398 | 57 | 54 | 25 |
Largely the same data, packaged differently. For "NASA", the News tab's first 9 results were the 9 articles in the feed's first item, in the same order:

- Same articles and times. 54 of 57 News tab results were in the feed, 10 of them inside a cluster's description. For the 44 matched as items,
publishedAtandpubDatewere identical. - Similar order. For "electric vehicles" and "Federal Reserve", the News tab followed the feed's ranking. For "Nvidia", its first five results were feed items 40, 55, 36, 2 and 19.
- Similar freshness. Median ages of the News tab results and the feed's top 10 were within an hour of each other for three queries; for "Federal Reserve", 18.4 and 35.7 hours.
- Different links and text.
/newsanswered in 3.6 to 4.1 seconds withlinksPending: true; polling turned all 57 links into publisher URLs 32 to 49 seconds after the request (how link resolution works). 25 of 57 results had a snippet; the feed had none.
When to use News search results instead
Use the Google News RSS feed for what its terms allow: a personal feed reader that needs headlines only. Use News tab results as JSON when you need article URLs, snippets when Google shows them, publish times and domains as fields, or a paid service with documented errors and limits. That is what BestSERP's Google News API returns, at $0.50 per 1,000 searches after 3,000 free ones (pricing).
One caveat: when Google blocks the News tab for a request, /news falls back to the Google News RSS feed and marks those items origin: "google_news_rss" (/news reference). In our Google Alerts alternative tutorial, that happened for 7 of 16 requests on November 18, 2026: about 47 seconds each, news.google.com links, no snippets, "www." in domain, no ipCountry, and one item from 2024 despite a 24-hour filter. None of our requests for this post fell back. Check origin in your code:
import os
import sys
import time
import requests
API = "https://api.bestserp.com"
HEADERS = {"X-API-KEY": os.environ["BESTSERP_API_KEY"]}
def get_news(q, gl="us", hl="en"):
for attempt in range(3):
resp = requests.post(f"{API}/news", headers=HEADERS, json={"q": q, "gl": gl, "hl": hl}, timeout=130)
if resp.status_code in (429, 502, 503): # 429 is never charged, 502/503 are refunded: wait and retry
time.sleep(int(resp.headers.get("Retry-After", 10)))
continue
resp.raise_for_status() # 400-402: fix the request instead of retrying
data = resp.json()
request_id, deadline = data["requestId"], time.time() + 150
# 202: still running. linksPending: the final article URLs are on the way. Polling is free.
while (resp.status_code == 202 or data.get("linksPending")) and time.time() < deadline:
time.sleep(5)
resp = requests.get(f"{API}/requests/{request_id}", headers=HEADERS, timeout=130)
data = resp.json()
return data
raise RuntimeError(f"/news {q!r} still failing after {attempt + 1} attempts")
data = get_news(sys.argv[1] if len(sys.argv) > 1 else "NASA")
print("ipCountry:", data.get("ipCountry"), "| items:", len(data["news"]), "| credits:", data["credits"])
for item in data["news"][:5]:
via = "RSS fallback" if item.get("origin") == "google_news_rss" else "News tab"
print(item["position"], item.get("publishedAt"), item.get("domain", "-"), f"[{via}]",
"snippet" if item.get("snippet") else "no snippet", "|", item["link"][:60])
Our run at 07:16 UTC took 33 seconds, most of it waiting for the final links (the final version, with its 150-second limit, printed the same lines at 07:21 from the cache, and took 33 seconds on a fresh query):
ipCountry: us | items: 10 | credits: 1
1 2026-11-17T11:00:00+00:00 latimes.com [News tab] no snippet | https://www.latimes.com/science/story/2026-11-17/nasas-voyag
2 2026-11-17T22:20:38+00:00 npr.org [News tab] no snippet | https://www.npr.org/2026/11/17/nx-s1-6043115/nasas-voyager-1
3 2026-11-17T17:00:00+00:00 space.com [News tab] no snippet | https://www.space.com/space-exploration/voyager/nasa-wants-y
4 2026-11-17T21:38:17+00:00 wltx.com [News tab] no snippet | https://www.wltx.com/article/news/nation-world/nasa-voyager-
5 2026-11-17T21:00:00+00:00 wcvb.com [News tab] no snippet | https://www.wcvb.com/article/light-day-voyager-1-visualized/
Every link is the publisher's URL. Since our 06:52 request, the Los Angeles Times article had moved from 3 to 1.
Common problems and fixes
| Problem | Cause | Fix |
|---|---|---|
| A topic or place feed is empty | Named topics and places answer 302; the client did not follow it | Follow redirects, or request the /rss/topics/ URL the redirect gives |
| Links do not reach the article | Item links are Google News IDs; plain HTTP gets a JavaScript page | Open them in a browser, or use a source with final URLs such as /news |
| Months-old items in a search feed | Feeds are sorted by relevance, not date | Add when:1d to the query |
| The same story twice | Clusters and overlapping feeds | De-duplicate on guid, then on the headline without the suffix |
/news items with news.google.com links |
The News tab was blocked; the RSS fallback answered | Check origin; retry later if you need final URLs |
How we tested
On November 18, 2026, 06:50 to 07:21 UTC, we made 21 requests to news.google.com/rss, at least 10 seconds apart: 20 with a desktop Chrome User-Agent, then the reader above with its own User-Agent at 07:21 (200, 92 items). The 20 were 11 feeds, 4 redirects (not followed automatically), 3 requests for 2 article links and 2 repeat fetches with the reader above. Every answer was 200 or a redirect to another news.google.com/rss address. The /news side was 7 BestSERP requests (gl: "us", hl: "en"), each checked for ipCountry. Results were matched by headline (85% similarity or more, without the publisher suffix). Raw feeds, responses and scripts are in the post's data folder.
Next steps
- For news inside a wider monitoring job, build a Google Alerts alternative in Python.
- For news next to shopping and local results, start with our guide to local and news search data.
- For JSON with final links, read the /news reference and try the Google News API.
- To give a language model fresh news with sources, see RAG with web search.