Wenxin Yiyan's 10-Tier Source Authority System — Where Does Your Brand Rank?

For most of 2026, the conversation around Generative Engine Optimization in China has been about whether AI search matters and how much visibility it can deliver. What has been missing — until late July — is the precise answer to the question every brand actually needs answered: "What does Baidu's AI consider a credible source?"

On July 26, a detailed analysis of Wenxin Yiyan's source authority architecture went public on Chinese GEO forums, drawing from platform documentation, search behavior testing, and the underlying logic of Baidu's recently upgraded Master Agent engine. For the first time, the system has been mapped into an explicit, 10-tier hierarchy. And it changes how overseas brands should think about their Baidu presence.

🪜 The 10-Tier Source Authority Hierarchy

Wenxin Yiyan does not treat all sources equally. It ranks them on a ladder where each step upward produces a geometric — not linear — increase in citation probability.

Tier 1 — Maximum Authority (Government & State Media): Government websites (gov.cn domains), state news agencies (Xinhua, People's Daily, CCTV), and peer-reviewed academic journals. Content at this tier is treated as ground truth and cross-referenced to validate lower-tier sources.

Tier 2 — Platform-Verified Authority (Baidu's Own Knowledge Graph): Baidu Baike entries and Baidu Knowledge Graph entities. This is Baidu's internal identity layer — when the AI needs to confirm that a brand or entity exists and is legitimate, it checks here first.

Tier 3 — Verified Commercial Authority (Official Corporate Websites): Enterprise websites with ICP filing, Schema.org structured data markup, and consistent entity information. A properly structured official website with machine-readable brand data sits at this level.

Tier 4 — Major News Media: Authoritative news outlets beyond state media — including mainstream portals and nationally recognized commercial news platforms. This is the news tier that provides the third-party citation signals for EEAT scoring.

Tier 5 — Baidu Ecosystem Content: Baidu Zhidao (Q&A), Baidu Wenku (documents), and Baijiahao (Baidu's publishing platform) — especially verified accounts with original content. Content published here carries native Baidu ecosystem signals.

Tier 6 — Major Portals & Vertical Media: Well-known industry portals, sector-specific authoritative publications. Provides niche authority signals within specific verticals.

Tier 7 — Regulatory & Legal Sources: Listed company announcements, legal documents, patent filings. Hard, verifiable data that crosses source boundaries.

Tier 8 — Third-Party Platforms: Zhihu, CSDN, Xiaohongshu, and other user-generated content platforms. Widely read but individually low-weight unless corroborated by higher-tier sources.

Tier 9 — E-Commerce Data: JD.com, Tmall product pages and reviews. Relevant for purchase-decision queries but carries limited standalone authority.

Tier 10 — Baseline (Self-Media & Forums): Personal blogs, forum posts, unverified self-media accounts. These sources are rarely cited individually and primarily serve as corroboration material when higher-tier sources already exist.

The critical insight: the jump from Tier 10 to Tier 8 is meaningful. The jump from Tier 8 to Tier 5 is substantial. The jump from Tier 5 to Tier 3 — or ideally Tier 2 — is the difference between being invisible and being the default answer.

🔍 The Triple Verification System — How the AI Judges What to Trust

The 10-tier hierarchy is not a static ranking. It is enforced by a three-stage verification process that runs continuously.

Stage 1: Multi-Source Identity Verification. The AI cross-references the entity claiming to be the source against multiple independent databases. It verifies whether the website's ICP filing matches the claimed organization. It checks whether the Baidu Baike entry for the brand is internally consistent with the information on the official website. If identity signals conflict across platforms, the source drops in the hierarchy — sometimes by multiple tiers.

Stage 2: Filter-Before-Use Cross-Validation. Before citing any data point, the AI checks whether the same fact appears in at least one higher-tier source. A claim made only on a Tier 8 platform with no Tier 1-5 corroboration is treated as unverified and typically excluded from the AI's synthesized answer. This is where the EEAT framework — Experience, Expertise, Authoritativeness, Trustworthiness — becomes concrete: without cross-platform verification, expertise is self-reported and authority is unproven.

Stage 3: Real-Time Monitoring & Correction. The AI continuously re-indexes and updates its source evaluation. Stale content — even from high-tier sources — gets flagged. Content that was once accurate but is now outdated loses authority over time unless refreshed. This means that authority is not a one-time achievement but an ongoing maintenance process.

⚠️ Where Overseas Brands Land — and Why It's a Problem

Now apply this hierarchy to the typical overseas brand entering China.

• Baidu Baike (Tier 2)? Not available without a Chinese registered entity.
• Official website with ICP filing and Schema markup (Tier 3)? Requires a Chinese business license to obtain ICP filing. Without ICP, the website exists in a regulatory gray zone that the AI's Tier 1 verification stage will flag.
• Baijiahao publishing account (Tier 5)? Requires Chinese phone verification and business documentation.
• News coverage on Chinese media (Tiers 4, 6)? Requires PR infrastructure and relationships that most overseas companies do not have at entry.

Without these assets, the brand's digital presence sits almost entirely at Tiers 8-10: third-party user platforms, e-commerce product pages, and possibly scattered self-media mentions. At these tiers, the triple verification system has almost nothing to cross-reference, and the citation probability — already low — drops further with every data inconsistency across limited sources.

This is not a ranking disadvantage. It is a structural exclusion from the AI's source pool.

🧭 How to Climb the Ladder — Three Moves That Work

The hierarchy makes the path forward unusually clear. Overseas brands need to move from Tier 8-10 to Tier 2-5, and the sequence matters.

Move 1: Establish Tier 2-3 Anchor Assets — Baike and Official Website. These two tiers are the foundation. Without Baidu Baike, the AI cannot verify the brand's existence at its primary identity layer. Without a properly structured official website, the AI cannot reconcile the Baike entry with an authoritative commercial source. Both are required before any lower-tier content strategy can compound in value.

Move 2: Activate Tier 5 Publishing — Baijiahao. Once the anchor assets exist, Baijiahao provides the publishing channel that feeds directly into Baidu's AI indexing pipeline. Content published here carries native ecosystem signals that third-party platforms cannot provide. The key is structured, data-rich content — not volume.

Move 3: Generate Tier 4/6 Cross-Referencing — News and Industry Coverage. The triple verification system rewards sources that appear consistently across multiple tiers. News coverage on recognized Chinese media portals — even industry-specific ones — provides the cross-referencing that turns a Tier 3 website into a Tier 2-verified entity in the AI's assessment.

🏢 The BPP Perspective

The 10-tier hierarchy is the most transparent map of Baidu AI's decision-making that has ever been publicly available. For overseas brands, it confirms what BPP has been saying throughout 2026: visibility in China's AI search is not a content problem. It is an identity infrastructure problem.

The structural barrier is explicit now. Tiers 2 and 3 — the tiers where AI visibility begins to become meaningful — require assets that cannot be created without a Chinese business entity. ICP filing requires a license. Baidu Baike verification requires a license. Baijiahao requires local documentation.

BPP exists to solve this exact gap. We provide overseas brands with the entity bridge that makes Tier 2-3 asset creation possible, the publishing infrastructure that activates Tier 5, and the strategic guidance that turns scattered Tier 8-10 mentions into a coherent, cross-verifiable presence across the hierarchy.

The 10-tier map tells you exactly where your brand is. The question that follows is the only one that matters: are you moving up — or staying where the AI cannot see you?