Yandex SERP & Knowledge Graph Scraper

Scrape Yandex web search across nine regional fronts and get the whole result page, not just ten blue links: organic rankings, sponsored rows, knowledge graph and AI answers, inline images and videos, product tiles, local companies, weather, and related searches. Choose page records for the familiar one-record-per-SERP shape or flat rows for rank tracking, ad intelligence and analytics. No login, API key, or proxy setup.

What it does

• Two modes: search by keyword, or parse Yandex SERP URLs you already have.
• Nine live fronts: yandex.com, .by, .kz, .uz, .com.tr, .com.ge, .az, .md, .tj.
• Rank tracking by region: 30+ named region presets plus a numeric regionId override, 19 interface languages, and correct past-day / two-week / month filters.
• The full SERP surface: organic rows with true slot and organic ranks, ads with click URLs and displayed domains, knowledge graph + AI answer text and sources, inline images, inline videos, products, local companies, weather and related searches.
• Page or row output: keep today's page records or switch to one flat row per item with recordType.
• Clean data: empty fields, arrays and objects are omitted at every depth, never sent as null.

Services

Service	What you get
Keyword search	Multi-query Yandex SERP collection with pagination, region, language, domain and recency controls
URL parsing	Fetch specific Yandex result pages from a list of URLs, including translated Yandex URLs
Rank tracking	Overall slot, organic-only rank, page number, query and region on every result
Ad intelligence	Sponsored title, snippet, click URL, displayed domain and sitelinks
Entity research	Knowledge graph attributes, AI answer body, cited sources and "see also" entities
Commerce research	Product titles, prices, old prices, currencies, shops, delivery options and ratings
Local research	Company names, categories, ratings, review counts, addresses, metro and opening hours
Media discovery	Inline image thumbnails and originals, video cards with duration, views, publish date and preview
SERP context	Weather card, related searches and result-count text

How it works

1. The actor builds a Yandex search URL from your query, domain, language, region and time filter.
2. It reads the server-rendered result page, using Yandex's own translated result view when a regional front challenges the request.
3. It parses every result block in page order and assigns ranks across organic and sponsored rows.
4. It removes duplicate URLs and duplicate block items, then stops at maxItems, an empty page, or the end of Yandex's 25-page window.
5. It pushes either one page record or one flat row per item. Page records keep the original dataset fields and add the new blocks beside them.

Input

Field	Type	Default	Description
mode	string	search	search by keyword or byUrls for existing SERP URLs
queries	array	["python programming"]	One or more search queries (mode=search)
serpUrls	array	–	Yandex SERP URLs, one per page (mode=byUrls)
domain	string	yandex.com	Regional front: .com, .by, .kz, .uz, .com.tr, .com.ge, .az, .md, .tj
language	string	en	Yandex interface/search language (19 options)
region	string	–	Named region preset such as moscow, almaty, istanbul
regionId	integer	–	Numeric Yandex region ID; overrides region
maxPages	integer	3	Pages per query, 1–25
startPage	integer	0	Zero-based first page, 0–24
maxItems	integer	100	Global unique-item cap, 1–10,000
timeFilter	string	any	any, day, twoWeeks, month
fixTypos	boolean	true	Let Yandex correct misspelled queries
includeAds	boolean	true	Include sponsored rows
includeBlocks	boolean	true	Include knowledge graph, AI answer, images, videos, products, companies, weather
includeRelatedSearches	boolean	true	Include related-search suggestions
outputFormat	string	pages	pages for page records, rows for flat items
failIfEmpty	boolean	false	Fail the run when nothing was collected

Example: rank tracking in Moscow

{
  "mode": "search",
  "queries": ["купить пластиковые окна", "пластиковые окна цена"],
  "domain": "yandex.com",
  "language": "ru",
  "regionId": 213,
  "maxPages": 5,
  "maxItems": 500,
  "timeFilter": "month",
  "outputFormat": "rows"
}

Example: audit a saved SERP URL

{
  "mode": "byUrls",
  "serpUrls": ["https://yandex.com.tr/search/?text=ikinci+el+araba&lr=87&p=0"],
  "includeBlocks": true,
  "includeAds": true,
  "outputFormat": "pages"
}

Output

Pages mode (outputFormat: "pages")

One record per SERP page. Existing fields are preserved; new fields appear only when the block exists.

• query, page, url, domain, lang, scraped_at
• total_results_text — the result-count sentence when Yandex publishes one
• organic_results[] — position (all slots), organicPosition, resultType, title, link, displayed_link, displayedDomain, host, snippet, favicon, thumbnail, date, sitelinks[], rating, reviewCount, price, priceCurrency, shopName, delivery[]
• ads_results[] — position, title, link, clickUrl, displayed_link, displayedDomain, url (only when Yandex exposes the destination), snippet, sitelinks[]
• knowledge_graph — title, description, image_url, source_url, attributes{}, label, answerText, sources[], seeAlso[]
• aiAnswer — label, text, sources[], imageUrl for standalone AI answer blocks
• inline_images[] — title, thumbnail_url, source_url, originalImageUrl, sourceHost, alt
• inline_videos[] — title, link, thumbnail_url, duration, durationSeconds, views, publishedAt, previewVideoUrl, sourceHost
• products[] — title, url, price, priceCurrency, oldPrice, rating, reviewCount, shopName, delivery[], imageUrl
• companies[] — title, url, rating, reviewCount, category, hours, priceLevel, address, metro, distance, features[], imageUrl
• weather — location, dateText, temperature, feelsLike, condition, wind, forecast[], hourly[], url
• related_searches[] — query, link
• search_metadata — requested_domain, served_domain, locale_switched, source_path, captcha_encountered, pages_fetched, results_found, region_id, region_name
• pagination_info — current_page, max_pages_set, has_next_page, results_per_page, start_page, source_exhausted

Rows mode (outputFormat: "rows")

One record per item with recordType set to organic, ad, knowledgeGraph, aiAnswer, inlineImage, inlineVideo, product, company, weather, or relatedSearch.

Every row carries query, page, domain, language, regionId, regionName, sourcePath, scrapedAt, and the fields that belong to its type: position, organicPosition, resultType, blockType, title, url, clickUrl, displayedUrl, displayedDomain, host, snippet, thumbnail, favicon, sitelinks[], rating, reviewCount, price, priceCurrency, oldPrice, shopName, delivery[], duration, durationSeconds, views, publishedAt, sourceHost, imageUrl, originalImageUrl, sources[], attributes{}, plus weather and company details when relevant.

Usage cost

You pay Apify's standard actor start and dataset-item pricing for this run. There is no proxy, API key, or external data subscription to configure, and the actor never asks you to supply credentials.

FAQ

Do I need a proxy or a Yandex account?

No. The actor uses a working connection path automatically and needs no login, cookie, or API key.

Why do some fields disappear from a record?

Empty values are omitted on purpose. If a page has no knowledge graph, no ad, no price, or no video, that field is absent rather than null or an empty array.

Why does served_domain differ from the domain I selected?

Yandex sometimes serves an international or regional entry point from another front, and .com is commonly served from .ru. The requested and actual fronts are both reported in search_metadata.

Why do I get fewer results than I asked for?

Yandex serves at most 10 results per page and 25 pages per query, so one query tops out around 250 results. Split a large job by region, time window, or query variant to collect more.

Are sponsored results included?

Yes, when includeAds is true. Each ad has a click URL and a displayed domain. A final destination URL is included only when Yandex exposes it; the actor never invents one.

What is the difference between position and organicPosition?

position counts every result slot in page order, including ads. organicPosition counts only organic rows. Use the latter for classic rank tracking.

Can I scrape a URL I already have?

Yes. Use mode: "byUrls" and paste one or more Yandex result URLs. The actor reads the query, region, page and time filter from each URL.

Are old saved tasks still supported?

Yes. Legacy keys such as searchQueries, yandexDomain, lang, lr, noReask, old time-filter names and the old proxy field are accepted and mapped automatically. They are not shown in the input form anymore.

Is this affiliated with Yandex?

No. This is an independent third-party actor that reads publicly available Yandex search pages.

Notes

• Yandex serves 10 organic results per page; page numbers are zero-based in the input and one-based in the output.
• maxItems counts unique items across all queries. In pages mode it limits the items inside page records; in rows mode it limits rows.
• regionId wins over region. Use it when you need an exact Yandex region rather than a named preset.
• language should match your query language to keep result text in the original language.
• Empty values are omitted at every depth. Arrays and nested objects are never sent as null, "", [] or {}.
• The actor stops on maxItems, an empty page, or the end of Yandex's page window. It never stops on a wall-clock timer.

Limitations

• Vertical search is separate. This actor returns the inline blocks that appear inside the web SERP. Dedicated image, video, news, and maps search belong to the separate Yandex vertical actors.
• Inline blocks are conditional. Yandex decides per query, front and layout whether to render image, video, product, company or weather cards. The actor emits a block only when the served page contains it; an absent block is not an error and is never synthesized.
• Ad destinations are sometimes hidden. Yandex often keeps only a click-tracker URL. clickUrl and displayedDomain are always honest; a direct destination is added only when Yandex reveals it.
• Dates are conditional. Web results rarely expose a publication date, so date appears only when the page shows one. Video publish dates are normalized to ISO when they can be parsed.
• Default region is context-sensitive. When region and regionId are empty, Yandex chooses a region from the query and request context, so two runs can rank differently. Set regionId (for example 213 for Moscow) when you need reproducible rankings.
• Deep pages lose relevance. Pages 11–25 are available, but Yandex's ranking gets less precise beyond the first ten. Use smaller page windows when relevance matters more than volume.
• Translated result text. When Yandex serves a regional front through its translated view, the original language is preserved when language matches the query language. A language mismatch can produce translated snippets.
• Localized labels. Some SERP labels (for example weather conditions and company categories) stay in the language Yandex renders for that front; the actor does not machine-translate them.