Baidu SERP API: Baidu search results as structured data

This Baidu SERP API turns any Chinese or English keyword into organic Baidu search results you can use straight away: one row per result with the title, the real destination URL, the snippet, the source site and the date where Baidu prints one. Use it to track rankings in China, research a market before you enter it, or feed Baidu results into your own tool.

Chinese and English keywords both work, and they return genuinely different result sets, so search in the language your audience uses.

What you get from the Baidu SERP API

• Real destination URLs. Every Baidu result links through baidu.com/link?url=…, a redirect that says nothing about where it goes. Baidu also writes the destination onto the result's own container, so url is where the result actually points and baidu_url is the redirect a browser would open. No extra request per row, no unresolved links in your export.
• Baidu's own properties, flagged. Baike, Zhidao, Tieba, Wenku and the rest appear throughout the organic list. They are real results and dropping them would misrepresent the page, so they stay, with is_baidu_property set to true. Filter on it for third-party pages only. The ratio is itself the finding: a keyword where Baidu's own surfaces take most of the page is a keyword you cannot rank on.
• Continuous rank. With several pages per keyword, rank keeps counting across the whole walk, so page two starts where page one stopped. That is what you want when you check where a domain sits.

Baidu SERP API use cases

• Rank tracking in China. Run your keywords on a schedule and join the exports on keyword and rank to see movement of your domain and of competitors.
• Market research. Check what Chinese searchers see for a product category before you enter the market, and which sites and Baidu properties dominate it.
• Content and SEO audits. Collect titles, snippets and sources for a keyword list to see how results are worded and how many are dated.
• Data pipelines. Use the dataset API as a Baidu search API for dashboards, spreadsheets or LLM research tools, in JSON, CSV or Excel.

How this compares with other options

The official Baidu search is a website, not a data feed, and the commercial SERP APIs that cover Baidu bill per search or by monthly plan. This actor charges per delivered result with a small start fee, needs no account with the engine and no API key of your own, and keeps the output shape stable. For other engines see the Google, Bing, Yahoo and DuckDuckGo SERP actors; for the Chinese market beyond search, see the e-commerce actors.

There is usually no result count

total_results is null on most pages, and that is Baidu rather than a gap here. The current layout prints "found the following results" with no number at all; only the older layout gives a figure, and where it does you get it.

Paging

Pages per keyword walks further into the results. rank is continuous across the whole walk, so page two starts where page one stopped rather than back at 1, which is what you want when you are checking where a domain sits.

Results per page accepts 10 to 50. Baidu often returns a few more than asked and sometimes fewer, so trust the rows rather than the request.

Errors

Code	Meaning
blocked	No attempt got a results page back
no_results	The page was served and matched nothing
bad_input	No keyword was given

blocked and no_results are worth telling apart here more than on most targets: an empty page from Baidu is far more often about where the request came from than about the keyword. Raise Attempts per page if you see blocked, and try the keyword in Chinese before concluding nothing matches.

Baidu SERP Scraper input

The input requires keywords. The initial query is 价格监控, and an English query used with this actor is price monitoring. Enter each query on its own line. Leading and trailing whitespace is removed, and blank lines are ignored. Repeated keywords remain separate searches, so you can compare their results without losing entries from your requested sequence.

pages defaults to 1 and accepts 1 to 10 pages per keyword. resultsPerPage defaults to 10 and accepts 10 to 50. These values describe how far the actor walks, while maxResults limits the entire dataset across all keywords. Its default is 200 and the maximum is 5,000. attempts defaults to 4 and accepts 1 to 15 attempts for a page that cannot be read. Increasing the result limit does not guarantee that Baidu has that many distinct results.

An input used for a real run is:

{"keywords":["price monitoring","跨境电商"],"pages":2,"maxResults":35,"attempts":6}

Example Baidu SERP API output

A real row from a run with the keyword price monitoring:

{
  "title": "Price Monitoring | Official Portal of the Department of Agriculture",
  "url": "https://www.da.gov.ph/price-monitoring/",
  "baidu_url": "http://www.baidu.com/link?url=-b8BolvSn65xzsjAKxZooY65TPhH7b0JeWsWMzkx-5RbkDJZbd6SXCFPVY-i5iFPeuBQe_qMSyOGTPfK5IV3sq",
  "display_host": "da.gov.ph",
  "source": "菲律宾农业部官网",
  "snippet": "home price monitoring price monitoring weekly average retail prices 2026 weekly average prices date file size september 21-27, 2026 138 kb september 14-20, 2026 126 kb september 7-13, 2026 138 kb august 31-sept 菲律宾农业部官网",
  "published": null,
  "is_baidu_property": false,
  "keyword": "price monitoring",
  "page": 1,
  "rank": 2,
  "total_results": null,
  "position": 2
}

Output and exports

The default dataset contains one row per organic search result. Each row identifies its keyword, page, continuous rank within that keyword, and position within the whole run. The title, destination url, baidu_url, display_host, source, snippet, published, is_baidu_property, and total_results retain their existing names.

Missing source information is returned as null. This is particularly common for publication dates and Baidu's total-result estimate. A result without an explicit destination can still have a useful Baidu redirect link. A source label is the text Baidu prints for that result; it is separate from the hostname parsed from the destination URL.

Download the dataset in JSON, CSV, or another export format supported by Apify. JSON preserves nulls and booleans; spreadsheet tools may display those values differently. Keep the keyword and rank columns when joining this export to a previous run. Search results change over time, so use the URL alongside the title when comparing individual pages.

The key-value store contains a summary with the number of results, distinct hosts, results on Baidu properties, dated results, and any collection errors. If a search is unavailable, the errors record explains which query failed. A partially completed search retains the results that were actually read. An empty search produces an empty dataset and a successful run with an explanation; it does not invent results to fill the requested limit.

Pricing

Pricing is pay per event: $0.002 per run start and $0.0025 per delivered result. The start event applies even when the query is blank or Baidu returns no readable results. The result event applies only to rows actually written to the dataset. An empty dataset therefore has no result charges.

For non-paying Apify users, delivery is limited to ten results per run and collection respects that limit. Those delivered rows retain the same event price. Paying users can request the larger limits described above. Apify's platform usage is separate from the actor's event prices; consult your run billing details for the full cost.

Frequently asked questions

Can I search in English?

Yes. Chinese and English keywords are accepted. The language of the keyword can change which websites Baidu returns, so use the wording your audience searches for when measuring visibility.

Why are there fewer rows than resultsPerPage?

Baidu controls the returned results. Entries without a usable title are skipped, duplicate destinations within a page are merged, and the global result limit can stop collection before the final page ends. The requested page size is not a promise of an exact row count.

Are Baidu-owned pages excluded?

No. They remain part of the organic result list and are marked with is_baidu_property. Filter that field when you need only external websites. Keeping the original entries also preserves the ranking you actually saw.

Does a missing published value mean the page has no date?

It means the result did not contain a date in the recognised format. Baidu can omit a date or use a different date representation. The snippet remains available for your own analysis, and the actor does not infer a publication date from unrelated numbers.