Skip to main content

Scraper API for Baidu

The ScrapeBadger Baidu API turns baidu.com into a clean JSON feed: the web SERP (with language and publish-date filters), the news vertical, image search with full-size image URLs and pixel dimensions, and search-box autocomplete. Baidu is China’s #1 search engine at roughly 60% market share — the single largest search surface outside Google, and the one that decides visibility for the Chinese-language web. It is also the search engine most Western SEO and market-research tooling simply skips.
All endpoints are GET, live under https://scrapebadger.com/v1/baidu/*, and authenticate with the X-API-Key header. Credits are charged per request (see the table below) and reported on the X-Credits-Used response header.

The real URL, not the redirect

Baidu does not link straight to a result. Every organic link is a baidu.com/link?url=... tracking redirect that only resolves to the destination when followed — which is why competing Baidu APIs hand you a baidu.com URL and leave you to make one extra HTTP request per result to learn where it actually goes. ScrapeBadger returns both, on every result:
So a rank-tracking or domain-visibility job needs one call, not 1 + N — you can group results by real hostname straight out of the response.

Global access, no China proxy

Baidu is reachable from outside China but rate-limits and challenges foreign traffic aggressively. ScrapeBadger absorbs that: requests are served through its own exit pool with session reuse and challenge handling, so you do not need a China-based proxy, a mainland VPS, or an ICP filing. Call the API from anywhere.

Features

  • Web search — organic results with title, real url, snippet, source (the site name Baidu shows), date/date_at, thumbnail and Baidu’s own tpl template id. Up to 50 results per page.
  • Language filter — restrict to simplified (zh-cn) or traditional (zh-tw) Chinese, or leave it open with all.
  • Publish-date window — time_from / time_to as Unix timestamps, mapped onto Baidu’s own gpc date filter.
  • Related searches — Baidu’s query suggestions from the foot of the SERP, with their SERP URLs, for keyword expansion.
  • News vertical — articles with publisher, publish date and real article URLs, ordered by relevance or recency.
  • Image search — full-size image_url (Baidu’s decoded objURL), three Baidu-hosted copies (thumbnail_url, middle_url, hover_url), pixel width/height, format, and the source page (from_url, from_title).
  • Autocomplete — Baidu’s search-box suggestions for 2 credits, the cheapest way to expand a seed term into real Chinese queries.

Dates

Every dated result ships two fields:
There is deliberately no timestamp field. Baidu’s SERP carries no time-of-day, so a midnight epoch would be fabricated precision. Sort on date_at and keep date for display.

Credits

Quickstart

Pagination limits

total_results is Baidu’s own claim (parsed from 百度为您找到相关结果约N个), not a reachable count. Baidu’s SERP stops serving results past roughly page 76 regardless of the total it reports, so page is clamped to 1–76 — a request beyond the wall is rejected rather than billed for a structurally empty page.
  • num controls page size, 1–50 (Baidu’s own cap). Fewer, bigger pages is the cheaper way to depth: num=50 reaches 500 results in 10 calls.
  • Page offset is (page - 1) * num, so changing num between pages shifts the window. Keep num fixed for the whole crawl of one query.
  • Images are a fixed 30 per page, page 1–50.
To go wider than one query’s page wall, expand the seed instead of paging deeper: run /autocomplete (1 credit) and the related_searches off the first SERP, then search each term and merge on url. Cheaper coverage than paging to 76.

Character encoding

Queries are Chinese in practice, so URL-encode the query as UTF-8 and let your HTTP client do it (URLSearchParams, requests params=, the SDKs). Every response is UTF-8; snippet and title come back with Baidu’s own punctuation (,, 、) intact rather than transliterated. Latin queries work too — Baidu serves a mixed SERP for coffee machine — but Chinese terms return meaningfully richer results.

Errors

Baidu serves a challenge page with HTTP 200. Those are detected by result- container marker, not status code, and surfaced as 422 so a challenge is never billed or mistaken for a genuinely empty SERP.

Run this asynchronously

Calls on this API can run in the background, so you don’t hold a connection open while they work. Add one of these to the call:
An async call is billed once, when it completes. See Async requests and webhooks for webhooks, signatures and the SDK helpers.
ScrapeBadger is an independent tool and is not affiliated with, endorsed by, or sponsored by Baidu. “Baidu” is a trademark of Baidu, Inc., used here only to describe the platform this scraper is designed to work with.