Start with the entity, not a file
State the official organization name, website, category, products or services, markets and contact route consistently. Explain relationships between the company, brands, product families and legal entity. If a similar name exists, add a short factual disambiguation statement.
This identity layer should be visible in normal HTML and consistent with appropriate Organization structured data. Structured data can clarify meaning, but it cannot rescue a page whose visible content is vague or contradictory.
Build pages around real decisions
Company facts
Who the organization is, what it does, where it operates and how to verify it.
Products and services
Capabilities, specifications, use cases, relationships, versions and limitations.
Comparison criteria
Explain how buyers should evaluate alternatives without manufacturing claims about competitors.
Evidence
Research, certifications, policies, cases and primary documentation with dates and scope.
Support
Compatibility, implementation, warranty, policies and questions that affect a decision.
Current commerce
Where relevant, separate changing price, availability and market policy from stable product facts.
Make each important claim usable
A useful claim states the subject, action or capability, applicable market, date or version, evidence and limitation in nearby text. Keep the canonical source at a stable URL. When a fact changes, update the source and preserve an internal version record rather than letting conflicting copies remain live.
Use headings that match the question, short direct answers near the beginning, descriptive internal links and meaningful page titles. Images and diagrams can support the explanation, but essential facts should also exist as text.
Technical and machine-readable layer
- HTTPS, successful responses, indexable canonical pages and genuine 404s
- Server-rendered titles, headings, main text and internal links
- Accurate canonical and language alternates without unnecessary duplicates
- XML sitemap and crawler policy that match the intended public surface
- Relevant structured data that agrees with the visible page
- Optional Markdown counterparts and llms.txt maps for agent-oriented discovery
- Automated tests for broken links, missing metadata and content drift
Machine-readable content maps are supplementary routes. They do not replace crawlability, indexing, semantic HTML, structured data or useful content.
A practical publication gate
- Approve factsConfirm source, owner, market, language, validity and public permission.
- Generate the pageWrite for the human decision first, then create the machine-readable counterpart.
- Validate parityCheck that HTML, Markdown and structured data do not make different claims.
- Publish and verifyRecord the final URL, content hash, release time, crawl response and index status.
- RetestUse the frozen question set to observe whether the intended representation changed.