Naver News Scraper

Fast, structured extraction of public Naver News search results. Version 3 uses direct HTML requests and Cheerio instead of launching a browser, so normal searches complete in seconds while preserving Naver pagination, proxy support, sessions, retries, sorting, date controls, and safe failure handling.

Input modes

• query: one Naver News search.
• queries: multiple searches. Records are selected round-robin so one query cannot consume the full global limit.
• startUrls: public https://search.naver.com/search.naver URLs containing where=news and a non-empty query.

These modes may be combined. Duplicate article IDs are removed; an article found by multiple searches receives matchedQueries and queryPositions.

Useful controls include sortBy (relevance, newest, or oldest), startDate, endDate, categories, publisher include/exclude lists, titleContains, pagination limits, concurrency, retries, timeouts, and Apify/custom proxy configuration.

{
  "queries": ["인공지능", "로봇"],
  "sortBy": "newest",
  "maxItems": 20,
  "maxPages": 2,
  "proxyConfiguration": { "useApifyProxy": true }
}

GOOGLE_SERP is intentionally rejected because that proxy group supports Google HTTP requests, not Naver.

Output

Records contain only observed or deterministic fields. Core fields include:

• stable id, articleId, and pressId;
• publisher article url and naverUrl;
• title, snippet, publisher, publisher profile/domain;
• ISO publishedAt, original publishedAtRaw, and date precision;
• Naver section/category when present;
• direct publisher image and publisher logo URLs when present;
• highlighted search terms;
• query, per-query and global positions, page, search URL, sort order, locale, and scrape timestamp.

Optional values are omitted when unavailable—no fabricated favicon, zero-count, false-status, placeholder, null, empty string, array, or object fields are emitted. Relative publication labels such as 2일 전 are converted to ISO timestamps and marked approximate.

Local validation

npm ci
npm test
apify run --purge --input-file qa-inputs/local-single.json
npm run validate:dataset
npm run audit:dataset

Additional bounded inputs cover multi-query fairness, pagination, filters, no-results, invalid input, and cloud proxy validation.

Blocking and safety

The Actor uses a consistent desktop HTTP profile, cookie-persistent sessions, bounded retries, and optional Apify/custom proxies. HTTP status and content type are validated before parsing. Recognized no-result pages produce an empty dataset. Access-denied, challenge, or unexpected markup fails closed and is never stored as data. The Actor does not solve or bypass CAPTCHAs, authentication, paywalls, or access controls.