One Strategy Per Platform: The "Platform Source-Preference Matrix" and Four Source-Compliance Red Lines for Pharma Brands

No "one content fits all platforms"

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

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

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"

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

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

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)

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.