--- title: "Content Authority Observable: Meta + OpenAI Co-Dominant Consumption and the Commerce Routing Chain" subtitle: "First-party bot logs confirm 97.3% content-read ratio with Meta + OpenAI at 88.6% combined — the same week Meituan and JD issued GEO tenders demanding ≥98% semantic accuracy" date: 2026-04-20 quantum_uid: "QID-GEO-CONTENT-AUTHORITY-OBSERVABLE-2026-04-20" tags: ["GEO", "AEO", "crawler-intelligence", "Meta", "OpenAI", "commerce-routing", "China-GEO", "Meituan", "JD", "bot-engagement-provenance", "content-authority", "generative-engine-optimization", "chinese-procurement", "semantic-accuracy"] author: "Protocol Maintenance Group" layout: "post" excerpt: "generativeenginesoptimization.org logged 97.3% content reads over 30 days, with Meta (69.9%) and OpenAI (18.7%) as co-dominant operators. This is not a discovery surface — it is source material. The same week, Meituan and JD simultaneously issued GEO procurement tenders demanding ≥98% semantic accuracy. That metric cannot be verified from the buyer side. This domain's operator consumption pattern is the verifiable proxy." --- --- title: "Content Authority Observable: Meta + OpenAI Co-Dominant Consumption and the Commerce Routing Chain" date: 2026-04-20 category: market-intelligence tags: GEO,AEO,crawler-intelligence,Meta,OpenAI,commerce-routing,China-GEO,Meituan,JD,bot-engagement-provenance summary: generativeenginesoptimization.org logged 97.3% content reads over 30 days, with Meta (69.9%) and OpenAI (18.7%) as co-dominant operators. This is not a discovery surface — it is source material. The same week, Meituan and JD simultaneously issued GEO procurement tenders demanding ≥98% semantic accuracy. That metric cannot be verified from the buyer side. This domain's operator consumption pattern is the verifiable proxy. --- # Content Authority Observable: Meta + OpenAI Co-Dominant Consumption and the Commerce Routing Chain Two things happened in the same week in April 2026 that the Chinese GEO procurement market does not yet know how to connect. First: Meituan issued its GEO procurement tender (registration deadline April 15). JD.com's home appliances division issued a parallel tender the same day. Both documents specify ≥98% semantic matching accuracy as a technical requirement. Second: thirty days of first-party bot logs confirmed that generativeenginesoptimization.org is being consumed as source material by two dominant LLM operators — Meta and OpenAI — at a combined rate of 88.6% of all known-operator traffic, with a 97.3% content read ratio. These two facts are the same phenomenon from different vantage points. ## The Consumption Profile From the strict known-operators dataset (850 hits, 30 days, unknowns excluded): | Operator | Hits | Share | Behavior | |----------|------|-------|----------| | Meta | 594 | 69.9% | 97.3% content | | OpenAI | 159 | 18.7% | 97.3% content | | Microsoft | 52 | 6.1% | — | | Google | 22 | 2.6% | — | | Amazon | 21 | 2.5% | — | | Anthropic | 2 | 0.2% | — | **Content ratio: 97.3%.** Of 850 known-operator hits, 827 were to content paths — not robots.txt, not sitemaps, not agent.json. Operators arrived knowing the domain exists and went directly to read. This is not discovery behavior. Discovery sweeps once or twice, maps the structure, and returns to verify. What 97.3% content reads over 30 days describes is an operator that has completed discovery and moved to ingestion. The domain is being consumed as a knowledge source. ## Why This Matters for Commerce The connection that procurement documents miss: what Meta and OpenAI ingest becomes training signal, inference context, and retrieval corpus. The domains they read at sustained cadence influence what their models cite, recommend, and route users toward. Chinese GEO platforms are now structuring commerce routing explicitly around operator ingestion. The confirmed routing chains as of April 2026: | Platform | Commerce Destination | Source | |----------|---------------------|--------| | Doubao (ByteDance) | Douyin shops + Doubao Wallet | AdQuan, April 2026 | | Qwen (Alibaba) | Taobao / Tmall / Alipay / Ele.me / Fliggy / AutoNavi | 36氪/EastMoney, Jan 2026 | | Yuanbao (Tencent) | WeChat Mini Programs + JD | IT之家, April 2026 | | DeepSeek | No commerce integration | — | | Kimi | No commerce integration | — | The path from content to commerce citation runs through operator ingestion. A domain consumed by Meta and OpenAI is feeding the inference pools that Doubao and Qwen query when routing commerce decisions. The pipeline is indirect but documented. Doubao's behavior is particularly direct: confirmed insertion of ads during meeting transcription, recommending vendors based on conversation content. No formal "sponsored GEO" program — but functionally identical to commerce-route injection. The ingestion pipeline is live. ## The Verified Proxy Problem Meituan's procurement standard (≥98% semantic matching accuracy) cannot be verified by the buyer. As 姚金刚 documented: the procurement criteria were generated by the vendors being measured, via AI search on those same criteria. The standard is circular. The procurement market is trying to measure something real — whether AI platforms actually read, cite, and route traffic from a domain. The metric it chose is unmeasurable from the buyer side. Bot engagement provenance — first-party log data showing which AI operators are reading, at what depth, how often — measures the same thing from the observable side. generativeenginesoptimization.org's 30-day profile is not a semantic accuracy score. It is evidence that Meta (594 hits) and OpenAI (159 hits) have decided this domain is worth reading continuously. That decision is the revealed preference the procurement standard is attempting to capture with a number that vendors can game. ## The Industry Standard Convergence On April 20, 2026, China's first industry group standard for "trustworthy GEO communication" (国内首部生成式引擎优化（GEO）可信传播团体标准立项) was officially initiated. The standard was launched specifically because the circular metric problem is now visible at procurement scale — Meituan and JD's tenders made it an enterprise-level problem, not just an academic one. The standard will need to resolve the verification gap. Bot engagement provenance — operator, depth, cadence, primary log evidence, SHA256-hashed exports, machine-readable schema — is the architecture that makes GEO claims auditable. The rail-taxonomy-1.0-beta.json schema at `agent-ads.org/schema/` was designed for exactly this: `audit_hash` on all context injection objects, `actor.operator` separation to distinguish platform from sponsor. The procurement standard is looking for a number. The observable is a behavior pattern over 30 days, confirmed by two independent LLM operators through their own crawl decisions. ## Optimization Target for This Domain Type A domain in content ingestion mode has a different optimization target than one in discovery mode. **What to maintain:** - Content depth — operators are reading, not just mapping; shallow content will drop the read ratio - Publishing frequency — sustained subscriber cadence (Meta-ExternalAgent at ~3-day return interval) means operators expect new content on schedule - Entity coverage — named entities in content get cited; gaps in entity coverage are gaps in citability - Semantic density — the actual signal in 97.3% content reads is that operators find the content worth processing; template content does not sustain this pattern **What to deprioritize:** Structural artifact polish (sitemaps, schema endpoints). Anthropic has 2 hits on this domain — infrastructure registration has not occurred at scale here. Content quality precedes infrastructure registration. **The co-dominant signal:** OpenAI at 18.7% (159 hits) is not a footnote. Two dominant LLM operators treating the same domain as source material is a stronger evidence claim than single-operator dominance. Redundant ingestion by independent operators signals that the domain is not niche. The content is being read across competitive inference stacks. ## MAU Corrections (DeepSeek Verification, April 20, 2026) Platform MAU numbers circulating in Western GEO analysis are materially wrong for the Chinese surface. Corrected figures from primary Chinese sources: | Platform | Previously Circulated | Corrected | Source | |----------|----------------------|-----------|--------| | DeepSeek | 194M | ~150-156M (Dec 2025/Feb 2026) | Multiple Chinese outlets | | Doubao | 174M | 227-408M (Feb-Mar 2026) | 中关村在线, 网易 | | Qwen | 100M | 203-322M (Feb-Mar 2026) | 21世纪经济报道 | | Kimi | ~10M | ~24.7M (Feb 2026) | Multiple | The corrected figures make the Chinese GEO surface larger, not smaller. Doubao at 227-408M and Qwen at 203-322M are not niche platforms — they are infrastructure at scale. Content consumed by Meta and OpenAI does not stay on the Western surface. --- *Evidence base: 30-day bot log analysis, workers_logs.db, strict known-operators dataset. CSV exports: `generative_domains_30d_summary_known_only.csv`. Machine-readable artifact: `generative_domains_30d_evidence_artifact.json`. Sources: DeepSeek verification sweep (2026-04-20), xAI verification C6.1 sweep (2026-04-20), Cycle 6 geo-aeo-tracker (2026-04-15). Platform routing confirmed via AdQuan, 36氪, EastMoney, IT之家.*