Editorial Standards

Last updated: 2026-09-29

This page states the quality bar that every China AI Hub page must clear before it is published. It is the editorial counterpart to the methodology page: methodology says how data is collected and verified; this page says what is allowed to be written around that data. The two link to each other because they operate together — data rules feed the writing, and the writing rules reference the data labels.

The quality gate

Before any page ships, it is checked against the following, in order:

1. No fabrication — every number and named fact traces to a cited source; missing values are marked "Not publicly documented" or "Unknown", never invented.
2. Source quality and freshness — sources are tiered (L1 first) and date-stamped.
3. Originality — editorial pages must pass the originality test; assembled-from-elsewhere content is rejected.
4. Completeness — entity schemas are filled or marked draft; a page over 50% null on core fields is not published.
5. Consistency — entity IDs resolve and versions agree across the site and the Data Hub.
6. Internal links — links present and semantically meaningful, with no dangling targets.
7. Structured data — Schema.org JSON-LD is valid.
8. SEO — title, meta description and canonical are set.

No winner verdicts

China AI Hub does not declare a "best" model, a "#1" provider, or an overall ranking. Comparisons are allowed to state a measurable difference for a specific metric (with source and date), and are otherwise phrased as trade-offs: "better suited to", "more relevant when", or "has an advantage in" a given workload — never "A > B" in general. This applies to entity pages, comparison pages and guides alike.

Four claim labels are required

Every factual or interpretive sentence falls into one of four labeled buckets — Official fact, Vendor-reported claim, Third-party evidence, or China AI Hub analysis — so the reader can always tell whose claim they are reading. Analysis is introduced as analysis and never presented as fact. The label definitions are on the methodology page.

Answer-block structure

Substantive pages lead with a direct answer rather than burying it. The standard shape is:

• Short answer — one sentence.
• Key facts — the sourced facts behind the answer.
• What this means — interpretation, labeled as analysis.
• What is uncertain — the honest gaps.
• Sources — the evidence.

This keeps pages useful to both readers and AI systems that extract direct answers.

Internal-link discipline

Links connect entities to their models, companies, agents, APIs and comparisons so the graph is navigable in one hop. Every link must point at a page that actually exists (verified before publishing), and anchor text must be descriptive rather than "click here". Orphan pages are treated as a defect and re-linked or removed.

Analysis formatting

Subjective interpretation is consistently prefixed (China AI Hub analysis: or Our interpretation:) and each conclusion stands on an evidence → reasoning → interpretation chain rather than jumping to a verdict. This keeps E-E-A-T stable: the site's authority comes from traceable reasoning, not assertion.

Division of labor with Methodology

To restate the split: methodology governs data (collection, verification, the 45/14 verification-status distribution, the eight-field evidence layer); this page governs prose (what may be written, how it is labeled, and how it links). Together with the author page, data policy, and corrections policy, they form the site's governance layer.