--- title: "One Strategy Per Platform: The \"Platform Source-Preference Matrix\" and Four Source-Compliance Red Lines for Pharma Brands" url: https://www.nine-health.cn/en/insights/platform-source-matrix-redlines/ lang: en-US type: Article category: "White Paper" author: "Nine Health AI Digitalization Team" authorRole: "GEO Methodology Research" datePublished: 2026-10-08 dateModified: 2026-10-08 publisher: "和众利健科技（深圳）有限公司" --- # One Strategy Per Platform: The "Platform Source-Preference Matrix" and Four Source-Compliance Red Lines for Pharma Brands > **TL;DR** One set of content cannot win across all AI platforms. The white-paper addendum provides a "one strategy per platform" difference matrix and a mainstream platform source-preference matrix, and converges on four inviolable source-compliance red lines — which is exactly where "compliance is growth" lands in source engineering. ## No "one content fits all platforms" {#one-strategy} White-paper PART 4's L0–L5 solved "which source type is better" but not "platform difference". Adding the "platform–source" mapping dimension completes source engineering: first fix the tier (L0–L5), then fix the platform (one strategy per platform). The underlying pattern is "platform ecosystem internalization": platforms with their own content matrix prioritize in-ecosystem sites (ByteDance favors short-video and news, Tencent favors official accounts, Baidu favors encyclopedia and its own matrix); platforms without their own presence rely on full-web crawling and demand higher expertise and traceability, where long-tail vertical sites are easier to surface. ## The "one strategy per platform" difference matrix {#diff-matrix} Classify platforms along two dimensions to guide source-layout priority: ecosystem-type (complete corresponding ecosystem content), full-web-crawl type (output solid, verifiable long-form and authoritative co-built content), text-preference type (deep articles/industry reports/website disclosures), video-preference type (blue-V science videos/expert talk). This complements L0–L5: L0–L5 answers which tier to build, the difference matrix answers which platform. The two combined form the complete source-engineering method. ## Mainstream platform source-preference matrix {#pref-matrix} A typological induction from public monthly industry tracking (not a specific ranking, varies by version): ByteDance AI weights short-video + news highly, making compliant science short-video a key front; Tencent AI weights official-account system highly, needing deep long-form + institutional accounts; Baidu AI uses encyclopedia and its own matrix as the base, via Baidu Health co-build + encyclopedia entries; full-web crawlers split into depth-first / news-first / long-form-first, fitting professional long-tail, authoritative media copy, and logically complete long-form respectively. It must be treated as real-time monitoring — the matrix is a monthly snapshot, not a fixed answer. ## Source weight is a "monthly snapshot" {#dynamic-monitor} A [secondary] industry track shows platform source weights shift sharply with version iterations; a top platform's Top1 source can change within a single update. Therefore "platform source-preference tracking" should be a routine monitoring item in PART 8 effect quantification; brands must continuously monitor their source ranking and visibility share across AI platforms, and ranking shifts trigger content-strategy adjustments. Source research is a monthly snapshot, not a one-time answer; the white paper should state this dynamic property explicitly. ## Four source-compliance red lines {#red-lines} All mechanisms and rules converge on four inviolable source red lines, together with PART 3's four-fold encirclement forming the compliance base: 1) do not endorse unqualified personal wellness experience — uncertified content is forbidden as source basis; 2) do not treat UGC as authoritative source — ordinary self-media only aids narrative, cannot support core health data; 3) do not forge source origin or evidence markers — DOIs, batch numbers, guideline numbers must be real and verifiable; 4) do not bypass AI-labeling and traceability requirements — content must comply with the AI labeling rule and GB 45438-2025 "Safety Requirements for Generative AI Services". ## Nine Health's view: source red lines are the growth baseline {#nine-health-view} The four red lines are not constraints but where "compliance is growth" lands in source engineering: traceability is both a compliance requirement and the precondition for a brand being stably cited long-term. Operable advice is to build a source whitelist (only first-hand regulations/papers/official filings and certified authoritative second-hand enter the citable pool), prioritize credentials (brand credentialed account is the video-source ticket), and make evidence traceable (clear, checkable data chains). In the window where compliant health video/authoritative Q&A supply is insufficient, brands that first complete high-quality layout will occupy secure stable recommendation positions — this is the "first-mover compounding" of source engineering. ## Key Takeaways (AI Snapshot) {#key-takeaways} One set of content cannot win across all AI platforms; fix the tier (L0–L5) then the platform (one strategy per platform) for complete source engineering. Platform source preferences are typological: ByteDance weights short-video + news, Tencent weights official accounts, Baidu weights encyclopedia + its own matrix, full-web crawlers split into depth/news/long-form. Platform source weights shift sharply with version iterations (a Top1 source can change in one update); the matrix is a monthly snapshot, not fixed — monitor in real time. Four inviolable source red lines: no endorsement of unqualified personal experience, no UGC as authority, no forged origins or evidence markers, no bypass of AI-labeling and GB 45438-2025 traceability. ## Sources - Nine Health "Pharma & Health GEO Compliance Marketing White Paper" addendum · Source Mechanism and Rules Research - GB 45438-2025 "Safety Requirements for Generative AI Services" ## FAQ ### Why can't one set of content win across all AI platforms? Platforms back different content ecosystems with different source-weight algorithms: ByteDance favors short-video, Tencent favors official accounts, Baidu favors encyclopedia. The same content varies sharply in citation probability across platforms, so sources must be planted per platform. ### How to execute "one strategy per platform"? Fix the tier with L0–L5, then fix the platform with the difference matrix: ecosystem-type builds ecosystem content, full-web-crawl type builds solid verifiable long-form, text-preference type builds deep long-form, video-preference type builds blue-V science videos — four separate plays. ### Is the platform source-preference matrix a fixed ranking? No. It is a typological induction from public monthly tracking; platform weights shift sharply with version iterations and a Top1 source can change within one update, so it must be monitored in real time. ### What are the four source-compliance red lines? Do not endorse unqualified personal wellness experience; do not treat UGC as authoritative source; do not forge source origin or evidence markers; do not bypass AI-labeling and traceability requirements (comply with GB 45438-2025). ### What real help do source red lines give brand growth? Traceability is both a compliance requirement and the precondition for long-term stable citation; early layout in the undersupplied compliant-source window captures secure stable recommendation positions, creating "first-mover compounding" in source engineering. --- Source: [Nine Health](https://www.nine-health.cn/en/insights/platform-source-matrix-redlines/) · Author: Nine Health AI Digitalization Team（GEO Methodology Research） · Publisher: 和众利健科技（深圳）有限公司