Guide · Baidu ERNIE

How do I get Baidu ERNIE (Wenxin Yiyan, Wenxiaoyan, Baidu AI Search) to recommend my product?

Updated 2026-11-12 · kagentx

Short answer

Baidu has not published the rules ERNIE, Wenxiaoyan or Baidu AI Search use to choose sources, nor any dedicated crawler for them; the Resource Platform's 2026-09-15 announcement says only that "AI traffic analysis" will be provided and the data is not formally enabled yet. What we can look up is Baidu's own search infrastructure: Baiduspider/2.0, Baiduspider-render/2.0, and the Resource Platform's general submission, fast crawl, sitemap and Robots tools. So what you can do now is: do not block Baiduspider, submit a single sitemap, put the body text in the HTML, build Chinese pages with their own URLs, and then ask three times yourself. None of our four sites' robots.txt files block it.

Which web pages does Baidu ERNIE's answer use?

Baidu does not say. We read the announcements and help pages on the Baidu Search Resource Platform and found no page explaining how Wenxin Yiyan, Wenxiaoyan or Baidu AI Search choose sources or which crawler they use. One thing is confirmed: on 2026-09-15 the platform announced "AI traffic analysis", saying it will provide data such as AI citations, impressions and clicks of site content in traditional search, but "the data service has not been formally enabled yet and complete data cannot be shown" (read 2026-11-12).

This shows Baidu treats "content being cited by AI" as data site owners should see, and that citation sits in the same system as the sites registered on the platform. As for which web pages ERNIE's answers draw from, Baidu does not say; claims in third-party articles (for example that Baidu Baike and Baijiahao are often cited) have no first-hand source we could find, so they are unverified. We also did not check Baidu's own pages for how the three names (Wenxin Yiyan, Wenxiaoyan, Baidu AI Search) relate, so we list them side by side.

Which crawlers has Baidu officially published?

NameOfficial statementSource (viewed 2026-11-12)
Baiduspider/2.0The User-Agent for both PC and mobile crawling is unified as Baiduspider/2.0, with http://www.baidu.com/search/spider.html in the UAResource Platform announcement, 2015-08-11
Baiduspider-render/2.0Rendering crawl: fetches a site's CSS, JavaScript and images to understand pages more accurately; please avoid blocking this UAResource Platform announcement, 2017-03-24
Baiduspider-imageImage crawling; Baidu's own robots.txt has a separate group for it, and the help page says to block image crawling with path rulesBaidu Help Center, www.baidu.com/robots.txt
Baiduspider-news, -cpro, -adsSeen only on third-party registry sitesNot verified, no first-party document
A dedicated crawler for ERNIE, Wenxiaoyan or AI SearchNo official statement foundNot verified

These announcements date from 2015 and 2017 and we found no newer official UA list, so go by your server logs for current UAs. For a name comparison see AI crawler user agents: a side-by-side table.

What should I know about Baidu's robots.txt rules?

Baidu's Help Center says the spider checks the root robots.txt before visiting a site; pages blocked by robots are not crawled, indexed or shown by Baidu, but if another site links to the page it can still appear in results, with a snippet taken from the other site's description.

RuleWhat the Baidu help page says
User-agentA robots.txt may have only one User-agent: * record; if you write User-agent: Baiduspider, the rules under that group apply
OrderThe order of Allow and Disallow matters: the spider matches from top to bottom and acts on the first match
Wildcards* and $ are supported
No robots.txtNo file or an empty file means the whole site is open to all search engines
ToolThe Resource Platform's Robots tool can validate and update it and show how it takes effect at Baidu; it supports content checks up to 48k

The help page has no date and its footer says 2011. The ordering rule differs from other engines (Google uses longest match), so the same robots.txt may behave differently at Baidu than at Google; check it once with the Robots tool.

How are our own four sites' robots.txt files set for Baiduspider?

Siterobots.txt rules (2026-11-12 curl, all 200)For Baiduspider
kagentx.comUser-agent * Allow: /; only /admin.html, /candidates.html and /api/ are blockedAllowed
mochistamp.comThe * group blocks /api/ (with Allow /api/content.js and /api/strings/ afterwards), /files/ and /admin.htmlHome page allowed; /api/content.js unverified (see below)
gammaslide.com* group blocks /d/, /decks, /account and /adminAllowed
pagent.kagentx.com* group blocks only /i18n/Allowed

None of the four sites names Baiduspider. Evaluated by text with Python's robots.txt parser, Baiduspider and Baiduspider-image are allowed on all four home pages. This is not a real Baidu spider visit, and we did not check whether Baiduspider appears in our server logs.

One thing about mochistamp.com deserves a separate look: it writes Disallow: /api/ first and Allow: /api/content.js afterwards. Under the ordering rule on Baidu's help page, the later Allow may not take effect for Baiduspider, while pages reference /api/content.js. We did not verify this with the Robots tool; unverified at the time of writing.

How do I hand URLs to Baidu? How do submission methods and quotas work?

ToolDocsSources
General submissionActively pushes data to Baidu to shorten the time the crawler needs to find links; three ways (API, manual, sitemap); inclusion is not guaranteedGeneral-submission page; Search College article (2020-08-06)
Fast crawlManual or API submission, crawled with priority, with no guarantee of inclusion or display; submitting uses up quotaSearch College article (2020-08-06); 2024-04-24 announcement: replaces "Fast Indexing" from 26 April
sitemapSubmission has a quota that differs by site; index-type sitemaps are not processed, and while one exists new files cannot be submittedSearch College article (2020-08-06)
Quota tighteningSitemap submission was switched off for sites under non-real-name accounts and for low-quality sites, and the daily API push quota was adjustedResource Platform announcement, 2023-09-22
Crawl diagnosisShows what Baiduspider actually fetched; 70 uses per week, showing only the first 200KBSearch College article (2020-08-06), numbers not checked one by one

On how to obtain fast-crawl entitlement, the 2020 article says it is granted automatically to VIP partners of the Resource Platform, or earned through platform events; we did not sign in to check current conditions; unverified.

What can an independent developer do now?

  1. 1. First confirm Baiduspider is not blocked

    Check robots.txt and your CDN's bot rules. Do not write a rule that lets in only Google. A reference version:

    # Let Baiduspider in; keep private paths out
    User-agent: *
    Allow: /
    Disallow: /admin/
    Disallow: /api/
    
    Sitemap: https://your-site.com/sitemap.xml

    For details on writing it see How do you write robots.txt for AI crawlers?.

  2. 2. Add your site on the Baidu Search Resource Platform and check with the Robots tool

    If you have rules where Allow comes after Disallow, use this tool to confirm how Baidu parses them.

  3. 3. Submit a sitemap, as a single file

    The official article says index-type sitemaps are not processed, so list URLs in a single sitemap; when you publish new pages, push them with the general-submission API. Inclusion is not guaranteed.

  4. 4. Put the body text directly in the HTML

    Baidu says only that rendering crawls fetch CSS and JavaScript, not how far execution goes, so price, features and the one-line introduction should be visible without scripts too; the reasons are in Do AI crawlers render JavaScript?.

  5. 5. If your target users ask in Chinese, build Chinese pages with their own URLs

    The Chinese page needs its own URL, a Chinese title and description; for multilingual writing see How do you write hreflang on a multilingual site? How does AI search treat it?. This is inference: Baidu does not say how ERNIE handles multilingual pages.

  6. 6. Ask in Baidu AI Search and Wenxiaoyan yourself, three times

    Ask the same question three times, note whether you are mentioned and which pages are cited; the method is in How do you measure AI visibility? Four metrics and a table you can follow.

For other Chinese models see How do I get Doubao and Kimi to recommend my product? and How do I get Qwen (通义千问) to recommend my product?; for the matching guide on the Korean market see How do I get Naver (네이버) to recommend my product?.

FAQ

Does Baidu Wenxin Yiyan's answer use Baidu Search pages?

Baidu does not say. The Resource Platform's "AI traffic analysis" announcement mentions citations and impressions of site content in AI, but does not explain how ERNIE fetches pages; unverified.

Does ERNIE have its own crawler?

We found no officially published name for a dedicated crawler; unverified. What Baidu publishes is Baiduspider/2.0 and Baiduspider-render/2.0.

What happens if robots.txt blocks Baiduspider?

The Baidu help page says blocked pages are not crawled, indexed or shown, but can still appear when other sites link to them, with a snippet from the other site's description.

Does submitting a sitemap guarantee inclusion?

No. Baidu says inclusion is not guaranteed; sitemaps also have a quota, and index-type sitemaps are not processed.

Are Baidu's robots.txt rules the same as Google's?

Not exactly. The Baidu help page says Allow and Disallow are matched from top to bottom and the first match is applied; check it with the Robots tool.

Can I use "AI traffic analysis" now?

The 2026-09-15 announcement says it is opening in stages and the data service is not formally enabled, so complete data cannot be shown. We did not sign in to check the current state; unverified.

Should I make Chinese pages specifically for Baidu?

If your target users ask in Chinese, build Chinese pages with their own URLs. This is inference; Baidu does not say how it handles multilingual pages.

How do I know whether ERNIE recommends me?

Turn on web search, ask the same question three times, and note whether you are mentioned and which pages are cited; Baidu does not publish its ranking method, and answers differ each time.

Sources

kagentx

Do AI answers
mention you?

An evaluation asks ChatGPT, Perplexity and Claude your buyers' questions (with web search on) and records who they recommend, which pages they cite and whether you are among them. Basic $1, results usually in 2–3 minutes.