Naver Web, Blog, Cafe & News Results Scraper

Naver SERP Harvester: Web, Blog, Cafe, News, Image & Video

Search Naver and collect the exact results a Korean reader sees: web documents, blog posts, cafe posts, news articles, images, videos, integrated result blocks and autocomplete suggestions. Every row carries the query that produced it, its position, the result link, a clean title and a normalized publish date whenever Naver prints one, plus a vertical-specific set of fields. No login, cookies or proxy setup is required.

What this actor does

• Eight modes: webSearch, blogSearch, cafeSearch, newsSearch, imageSearch, videoSearch, integratedSearch, autocomplete
• Batch queries: pass one or many queries; duplicates and blank entries are removed, and the same result found twice is emitted once
• Rank and provenance: every row has query, resultType, rank, page and the search page it came from
• Filters: relevance or newest-first sorting, date windows, paid-result control, and optional full blog/news content
• Clean output: exact destination URLs, normalized dates with a precision label, thumbnails, image dimensions, video duration and views, place details, and AI-briefing text and sources
• Empty fields are omitted — you will not see nulls, empty strings or empty arrays in the dataset

Output per result

Every normal row:

• recordType — searchResult; failed queries produce error
• resultType — web, blog, cafe, news, image, video, place, aiBriefing, ad or suggestion
• query, mode, rank, page
• title, url, resultUrl (same value), domain, sourceName, snippet
• thumbnailUrl
• date — the raw date text Naver prints
• publishedAt — ISO 8601 timestamp normalized to Korea time
• publishedAtPrecision — exact, date or relative
• publishedAtRaw — the date token used for normalization
• serpUrl, sourceUrl — the search page that produced the row
• scrapedAt — ISO 8601 collection timestamp

Mode-specific additions:

• Web: isAiSource when the result is flagged as an AI source
• Blog: blogId, logNo, isAd for paid placements
• Cafe: cafeName, cafeUrl, articleId, tags[], commentCount
• News: naverNewsUrl, pressUrl, officeId, articleId, section, sectionId, isRelated for cluster articles
• Image: imageId (also imgId), imageUrl, originalUrl, width, height, section, author
• Video: duration, durationSeconds, views, videoId, platform, profileUrl
• Place: placeId, category, rating, reviewCount
• AI briefing: aiBriefing with text, generatedAt and sources[] (each with title, URL, platform and publish time)
• Autocomplete: suggestion, plus answer when Naver returns an answer box
• Full content: content, contentLength, images[] when includeContent is on and the source exposes a body

A failed query adds one error row with recordType: "error", errorType, message, sourceUrl, scrapedAt and query.

Input

Example: newest blog posts with full bodies

{
    "mode": "blogSearch",
    "searchQueries": ["비비바", "서울 맛집"],
    "maxItems": 200,
    "maxItemsPerQuery": 100,
    "maxPagesPerQuery": 5,
    "sortBy": "date",
    "dateRange": "1m",
    "includeAds": false,
    "includeContent": true
}

Example: cafe discussion from the past week

{
    "mode": "cafeSearch",
    "searchQueries": ["제주도 여행"],
    "maxItems": 100,
    "sortBy": "date",
    "dateRange": "1w"
}

Example: news and image monitoring for a brand

{
    "mode": "newsSearch",
    "searchQueries": ["BTS", "삼성전자"],
    "maxItems": 300,
    "maxPagesPerQuery": 10,
    "sortBy": "date"
}

Example: integrated page with places and AI briefing

{
    "mode": "integratedSearch",
    "searchQueries": ["비트코인 전망"],
    "maxItems": 50,
    "includeAds": false,
    "includeSerpFeatures": true
}

Example: autocomplete ideas

{
    "mode": "autocomplete",
    "searchQueries": ["비비", "서울 맛집"],
    "maxItems": 20
}

Use cases

• Korean SEO and rank tracking — record which pages and blogs rank for a query, in relevance or newest-first order, with stable ranks and exact URLs
• Brand and news monitoring — batch a company or product name, restrict to a date window, and keep publisher, section, headline, snippet and publish time
• Content and RAG datasets — collect full blog and news bodies with authors, images and normalized timestamps
• Local and place research — pull the place pack from integrated results with ratings, review counts, categories and map links
• Visual and video research — collect images with original dimensions and videos with duration, views, channel and platform
• Keyword discovery — expand seed terms with Naver autocomplete suggestions

FAQ

Do I need a Naver account, cookies or a proxy? No. The actor collects public Naver result pages without login, cookies or any proxy configuration.

Why does webSearch return only a handful of rows? Naver's web-document section on the integrated page currently shows roughly 2–17 documents per query and does not paginate. Use integratedSearch to also capture places, ads and AI briefings, or the dedicated blog, cafe, news, image and video modes for deeper verticals.

How deep can the verticals go? It depends on the query. Broad Korean queries can return hundreds of blog, cafe, news, image or video rows; narrow or very recent queries return fewer. The actor stops when Naver runs out of results, when a page repeats, or when your maxPagesPerQuery / maxItems limits are reached. No rows are invented to fill a quota.

Are paid results included? Only when includeAds is on. Paid blog placements and integrated-search ads are marked isAd, and you can switch them off.

What does includeContent add? For blog and news rows it fetches the post or article body, inline image URLs, the author where available, and an exact publish time. Some posts are media-only or paywalled; those rows keep their card data and simply omit the missing content fields.

Why are some publish dates relative? Naver often prints labels such as 3주 전 or 11시간 전. Those rows keep the raw label in date / publishedAtRaw, get a computed publishedAt, and are marked publishedAtPrecision: "relative" so you can treat them differently from exact timestamps.

Why is there no cafe body or comment text? Naver does not expose cafe bodies or comment threads on public search pages. Cafe rows include the card fields Naver prints, including the comment count.

Why do some image links look like redirects? Naver's image tab returns signed bridge links for shopping and clip results. They are the official destination Naver serves at collection time; some merchant or short-link tokens expire later, so keep imageUrl / originalUrl when you need the media itself.

What is integratedSearch? It collects the blocks on Naver's main result page: web documents, paid ads, place cards and the AI briefing when Naver generates one. Place and AI rows can be turned off with includeSerpFeatures.

Can I search several queries at once? Yes. Each row records the query that found it, and a result found by two queries is emitted only once.

Is this affiliated with Naver? No. This is an independent actor that reads Naver's public search results; it is not endorsed by or affiliated with NAVER Corporation.