For two years, one question had no official answer: is Baidu's AI citing my site, and how often? The only figures available came from third-party vendors — including GEO service providers with a commercial interest in the number they reported.
That changed, quietly, in mid-September. Baidu Search Resource Platform (百度搜索资源平台) — Baidu's equivalent of Google Search Console — added an AI Traffic Analytics section. Most site owners learned about it from an in-platform notification rather than an announcement.
It is not fully live yet, and the data is limited. But the direction is unambiguous: AI citations are now an official site metric in China's largest search ecosystem.
📦 What Baidu shipped
Two modules appeared in the platform.
AI Traffic Analytics reports four core metrics:
- AI citation count (AI 引用量) — how many times a site's pages were cited in an AI answer.
- AI impressions (AI 曝光量) — how many times users saw a link to a cited page.
- AI clicks (AI 点击量) — how many times users clicked a cited page.
- AI click-through rate (AI 点击率) — clicks divided by citations.
Alongside these sits cited page count (引用页面量), plus two breakdowns: top keywords triggering AI citations and most-cited pages. The keyword view shows which queries caused your content to be cited, with citation, impression and click volumes attached.
Baidu's own wording is precise and worth quoting carefully: the module covers AI citation, impression and click data "for site content within the traditional search scenario." That framing matters — see the limitations below.
Search Mini-Tools (搜索小工具) is the second addition, structured in three parts: submission, a list of your tools, and a data dashboard. It is not a search-box plugin. It lets a site package scattered resources — Q&A, lookups, quotes, inventory, class schedules, service locations — into structured mini-tools submitted to Baidu. When Baidu detects matching search demand, it can call those capabilities directly instead of returning a list of pages.
🔍 From vendor estimates to official numbers
The significance is clearest by contrast.
Our September 10 analysis of ERNIE's citation behavior relied on a widely circulated figure: Baidu-ecosystem content accounts for about 81.7% of ERNIE's citations, with Baijiahao alone at 41.3%. We flagged that measurement as third-party, and noted that its source was a GEO vendor whose services include exactly the placement it was measuring — a commercial interest in the result.
That caveat does not vanish because Baidu shipped a dashboard. But one thing changes: a brand can now check a service provider's claims against Baidu's own data. For site-level AI citations inside Baidu's ecosystem, the black box has a window.
This connects to our September 11 piece on China's GEO compliance reckoning. The most common warning sign in that investigation was a vendor that would not provide an operations ledger, fact-verification records, or a change history. When official citation data becomes available, guarantees like "top-three AI mention, guaranteed" get easier to falsify — and harder to sell.
📖 How to read the four metrics
The metrics are more useful in combination than individually.
A high citation count with a low click-through rate means AI answers are using your content as source material, and users got what they needed without visiting. Baidu's guidance treats this as value, not failure: the citation itself is the outcome.
A high click-through rate means the reverse — the AI answer did not fully satisfy the user, who then came to your page with a specific question. Those pages are natural conversion entry points and deserve deliberate optimization.
For B2B brands, the top-keyword breakdown is the most actionable part. It shows which commercial queries trigger citations: a direct map of where your Chinese AI visibility already exists, and — by omission — where it does not.
🧰 Search Mini-Tools is the more interesting launch
AI Traffic Analytics measures. The mini-tools module acts.
For manufacturers, distributors and service businesses, the appeal is specific. A quote calculator, an inventory lookup, a service-location finder or a class schedule is exactly what a buyer would otherwise dig through pages to find. Submitting it as a structured tool means Baidu can invoke it when demand appears — moving a brand from "a page that might be cited" to "a capability that gets called."
This favors B2B over consumer brands, and it favors companies with something concretely structured to offer. It also raises the bar in a way that matters for overseas brands: structured submission requires Chinese-language data architecture and a verified Chinese presence.
🚧 What it does not solve
Four limits deserve equal billing with the launch.
- It is not fully live. The capability is rolling out in phases and, per Baidu's notice, the data service is not formally enabled — complete figures cannot yet be displayed. New modules typically carry reporting lag; data stabilizes in about one to two weeks. Blank dashboards during initial verification are stated to be normal.
- The scope is narrower than "AI." Baidu's wording is traditional search scenarios. This is not a complete picture of AI visibility across Chinese assistants — it is Baidu's own surface.
- It covers one ecosystem. Doubao, DeepSeek, Qwen, Yuanbao and Kimi publish no comparable official data. Cross-platform visibility still has to be assembled from manual testing and third-party tooling.
- Access requires site verification. Using the module means verifying domain ownership through Baidu Search Resource Platform. For an overseas brand without a China-accessible site or local entity support, that is a real barrier — the same barrier that runs through every layer of Chinese AI visibility.
🧭 What to do, and where BPP fits
Practically:
- Verify your site in Baidu Search Resource Platform first. Without verification there is no data, and without data there is no way to audit a vendor.
- Treat the keyword breakdown as a content map. It shows which commercial queries already trigger citations — build on what works before expanding.
- Watch click-through rate by page, not in aggregate. Low-CTR pages are citation assets; high-CTR pages are conversion assets. They need different treatment.
- Evaluate Search Mini-Tools for structured B2B offerings. Quotes, inventory, locations and schedules are strong candidates — if you can produce the Chinese structured data.
- Do not mistake one dashboard for cross-platform visibility. Baidu's data covers Baidu. Your Chinese buyers are spread across six assistants.
BPP works at the intersection this creates: Chinese-entity access and site verification without a Chinese license, structured Chinese content prepared for submission, cross-platform AI visibility across the six Chinese assistants, and citation data reconciled against what vendors claim.
Baidu just turned the lights on in one room. The rest of the house is still dark.
Key Takeaways
- Baidu now reports AI citation count, impressions, clicks and CTR for site content within traditional search scenarios.
- For the first time, a brand can check a GEO vendor's claims against Baidu's own data.
- Search Mini-Tools is the actionable half: structured quotes, inventory and locations that Baidu can invoke directly.
- Four limits: not fully live, "traditional search" scope only, Baidu ecosystem only, and it requires site verification.