Naver News Search Scraper - monitor Korean press coverage by keyword, day by day

Search Naver News (네이버 뉴스), Korea's main news portal, for any keyword and get one row per article: headline, press, age, snippet, thumbnail, the press's own link and the Naver News link. Give a date range and the actor walks it day by day, newest first, for press monitoring. Switch to relevance and Naver's related articles from other presses come as rows too, so coverage per story is one group-by away. Switch on Open each article to add the exact publish and update time, the sections and the full text.

It reads the same public pages anyone sees in a browser, in Korea's own news search and on the article page itself. No Naver account, no API key, no login, no cookies.

Read this first

• Naver's public search page shows 10 stories per page and has no second page. Without dates you get the 10 newest articles (sortBy newest) or the 10 most relevant stories with the related articles Naver lists under each (relevance: 10 rows for a keyword without related coverage and up to 50 for a busy one; 22 to 30 rows for the three busy keywords we ran on 2026-10-31 and 2026-11-01). With a date range the actor asks once per Korea-time day and returns the 10 newest articles of that day. A busy keyword is therefore sampled, not complete: for 스타벅스 on 2026-10-30 the 10 articles of the day were written between 19:00 and 23:38 Korea time, the last hours of that day. A quiet keyword, with fewer than 10 articles a day, returns every article Naver lists for that day; a day without an article gives no row and the run summary counts it in emptyDays.
• publishedTimeText is Naver's own text, a relative text (48분 전, 11시간 전, 1일 전, 3주 전), a date (2026.07.15.) or the print-edition position of the article (A23면 1단). The exact time (publishedAt, UTC) comes with Open each article.
• Only articles that Naver News hosts have a Naver News link and can be opened. On the newest-first pages we read on 2026-11-01 this was 0 to 9 of 10 articles, 3.7 on average over 29 pages, and in relevance mode about 9 of 10 stories (the other articles are only on the press's own site: their naverNewsUrl is null and, with Open each article on, their contentState is notOnNaver; they cost no extra request). The press's own link (articleUrl) is always there.
• Open each article costs one more request per article and the actor waits 2.5 seconds between any two requests, so a page of 10 articles with their full text takes up to about 30 seconds. Without it a search page takes a few seconds. A very large run (many keywords and days) may stop early at the run's time limit: it then keeps every row it already delivered and says time_budget in the run summary. Nothing is lost silently.
• Some Naver-hosted articles cannot be read as articles. Entertainment and sports articles that Naver moves to its own entertainment or sports sites answer with a redirect (the actor never follows it) or with a page of another layout: they keep their search row, with contentState unavailable and empty article fields. The same holds for an article Naver has deleted. Such a row is still delivered and counted; the run summary counts them in contentUnavailable (and the redirects among them in articleRedirects).
• E-mail addresses in snippets, headlines and article text, and phone numbers in snippets and article text, are replaced (best effort) with [email redacted] and [phone redacted]: addresses in common spellings (also name (at) host.com and Korean name골뱅이host.com) and Korean mobile and landline formats (010-1234-5678, 02-398-8585, +82 10 ...), also when several numbers stand side by side. Some articles end with a reporter's or desk address or phone line (in our test runs 2 of 14 articles held an address on 2026-10-31, one of them also a phone number, and 6 of 26 on 2026-11-01); it comes out as [email redacted] or [phone redacted]. Unusual spellings can slip through.
• A refusal from Naver ends the run. If Naver answers a search page with 403, 429, a redirect or a page that is not the news search page, or answers an article page with 403, 429 or another unexpected status, the actor stops, keeps every row it already collected and says blocked in the run summary. It does not retry, rotate or disguise itself.
• Not covered: comments, reactions, the article's images, the other Naver tabs (blog, cafe, shopping), articles beyond the 10 per keyword and day that the page shows, and the full related-news list behind a story's "view all" link (the page lists up to 4 related articles under a story; the link behind it counted 37 for the top 스타벅스 story on 2026-10-30). Not affiliated with Naver.

What you get

One row per article:

field	meaning
keyword, sortBy	the keyword you searched and newest or relevance
windowDate	with a date range: the Korea-time day the row was found for (2026-10-25); otherwise null
rank	position of the row on that search page (1 = first)
clusterRank, isRelated	the story the row belongs to (1 = first story of the page) and whether it is a related article of that story (true) or the story's main article (false); related articles appear in relevance mode only
title, snippet	the headline and the excerpt Naver shows (search highlights removed, contact details replaced); snippet is null for related articles. Naver cuts long headlines and excerpts and marks the cut with ... (about 1 headline in 4 in our runs on 2026-11-01)
articleUrl	the press's own link to the article
naverNewsUrl	the article's Naver News link (https://n.news.naver.com/mnews/article/001/0016411348?sid=102); null if Naver News does not host it
pressName, pressOid, articleAid	the press and, for Naver News links, the press number and article number (001, 0016411348)
publishedTimeText	Naver's text for the age or date, for example 11시간 전
thumbnailUrl	the article's picture on Naver's image server; null for related articles and articles without a picture
contentState	null when you did not ask for article pages; ok when the article page was read; unavailable when it could not be read; notOnNaver when Naver News does not host the article
publishedAt, modifiedAt	exact publish and update time, ISO UTC (2026-10-30T02:07:34.000Z); modifiedAt is null when the page shows no update time; with Open each article
sections	the sections the press filed the article under (["사회","경제"])
content, contentLength	the full text, one paragraph per line, contact details replaced; its length in characters
scrapedAt	when the row was collected

The key-value store also holds RUN_SUMMARY: complete, terminationReason (finished, max_results, charge_limit, time_budget, blocked, request_failed), rowsPushed, one entry per keyword (pages, rows, emptyDays, contentUnavailable, articleRedirects, notOnNaver), keywords left out beyond the first 50 (skippedOverLimit) and warnings. complete is false when the run ended before it could deliver what you asked for (time_budget, charge_limit, blocked, request_failed). If the platform restarts a run, articleRedirects can miss some redirects of the earlier part; every other number stays exact.

Input

field	default	meaning
keywords (required)	스타벅스	what to search for, one keyword or phrase per line, up to 50 per run (Korean or English); blanks and duplicates are dropped
sortBy	newest	newest (needed for date ranges) or relevance (10 stories with their related articles; dates are ignored)
dateFrom, dateTo	empty	YYYY-MM-DD in Korea time. With newest, every day from dateTo back to dateFrom is searched, newest day first. dateTo alone is refused; dateFrom alone ends today (Korea). At most 366 days. Leave both empty for the 10 newest articles. The dates stay fixed in a saved input, so a schedule repeats them: for yesterday's articles only, set the dates from your own scheduler or API call, or run without dates and drop repeats by articleUrl
maxArticlesPerKeyword	100	1-5000 rows per keyword (the default stops after 10 days of a date range, so raise it for a longer range)
includeArticleContent	off	open each article hosted on Naver News: exact time, sections, full text
proxyConfiguration	Apify datacenter proxy	the default is what this actor was tested with

Examples:

• The 10 newest articles about a brand: {"keywords": ["스타벅스"]}
• Three days, two keywords, 10 articles a day each: {"keywords": ["스타벅스", "삼성전자"], "dateFrom": "2026-10-29", "dateTo": "2026-10-31"} returns 60 rows.
• Coverage per story: {"keywords": ["스타벅스"], "sortBy": "relevance"} returned 22 rows (10 stories and 12 related articles) when we ran it on 2026-10-31; group by clusterRank to count the presses per story.
• The newest 10 articles with full text: {"keywords": ["삼성전자"], "includeArticleContent": true, "maxArticlesPerKeyword": 10}

Example row

{
  "keyword": "스타벅스",
  "sortBy": "relevance",
  "windowDate": null,
  "rank": 1,
  "clusterRank": 1,
  "isRelated": false,
  "title": "스타벅스, '탱크데이' 불매운동 후 광주 대형매장 첫 폐점",
  "articleUrl": "https://www.yna.co.kr/view/AKR20261030074300054?input=1195m",
  "naverNewsUrl": "https://n.news.naver.com/mnews/article/001/0016411348?sid=102",
  "pressName": "연합뉴스",
  "pressOid": "001",
  "articleAid": "0016411348",
  "publishedTimeText": "11시간 전",
  "snippet": "스타벅스코리아가 '탱크데이' 불매운동 후 처음으로 광주에서 대형 매장을 폐쇄했다. 30일 스타벅스코리아에 따르면 전남광주 동구 산수동 무등산국립공원 초입의 '스타벅스 광주지산유원지입구점'이 이날 문을 닫았다. 2017년 개점 이후 인근 거주 주민들은 물론 무등산 방문객 등도 많이 이용했던 지상...",
  "thumbnailUrl": "https://imgnews.pstatic.net/image/origin/001/2026/10/30/16411348.jpg",
  "contentState": "ok",
  "publishedAt": "2026-10-30T02:07:34.000Z",
  "modifiedAt": "2026-10-30T02:08:16.000Z",
  "sections": [
    "사회",
    "경제"
  ],
  "content": "'탱크데이' 파문에 깨지고 부서진 스타벅스 텀블러\n[연합뉴스 자료사진]\n(전남광주=연합뉴스) 정회성 기자 = 스타벅스코리아가 '탱크데이' 불매운동 후 처음으로 광주에서 대형 매장을 폐쇄했다.\n30일 스타벅스코리아에 따르면 전남광주 동구 산수동 무등산국립공원 초입의 '스타벅스 광주지산유원지입구점'이 이날 문을 닫았다.\n2017년 개점 이후 인근 거주 주민들은 물론 무등산 방문객 등도 많이 이용했던 지상 2층짜리 단독 상가건물의 대형매장으로 전날 영업을 끝으로 운영을 종료했다.\n스타벅스코리아는 상가 임대차 계약 종료로 해당 매장을 폐점한 것으로 알려졌다.\n스타벅스코리아는 5·18 46주년인 올해 5월 18일 텀블러 할인 행사에 1980년 5월 당시 계엄군의 무력 진압을 연상케 하는 '탱크데이' 표현을 사용해 물의를 빚으며 불매운동으로 번졌다.\n스타벅스는 5월부터 넉 달 연속 국내 커피브랜드 1위(신용·체크카드 추정 결제액 기준)를 다른 업체에 내줬고, 지난달 들어 추정 결제액 1천200억원대로 선두 자리를 되찾았다.\n[email redacted]",
  "contentLength": 534,
  "scrapedAt": "2026-10-30T13:44:55.000Z"
}

Good to know

• Press monitoring: run newest with the previous day as dateFrom and dateTo once a day, with maxArticlesPerKeyword 10 (or more if you list several days), and join on articleUrl.
• Coverage per story: relevance shows what Naver ranks first for a phrase and which other presses ran the same story (clusterRank, isRelated).
• Speed: one request at a time, a 2.5 second pause between requests. Through the proxy plan about 3 seconds per request, search page or article page (runs on 2026-10-31 and 2026-11-01).
• Cost control: you pay per delivered row, so maxArticlesPerKeyword is also a cost cap. The platform's own spending limit stops the run (charge_limit in the summary) and keeps the rows already delivered.
• Restarts: if the platform moves the run to another server, the actor continues where it stopped and does not deliver (or charge) a row twice.

Pricing

Pay per result, at the rate shown in the pricing panel of this page, plus a very small per-run start fee. You are charged only for the results delivered.

Related searches

naver news scraper, naver news search, 네이버 뉴스 크롤링, 네이버 뉴스 검색, korean news monitoring, press monitoring korea, naver scraper