Best Practices

Guidelines for authoring high-quality AI Page Manifests.

1. Write a clear purpose

purpose answers "why does this page exist?" in one sentence. It should describe the reader outcome, not repeat the title.

// Good
"purpose": "Explains how SMB websites deploy llms.txt and entity.json via WordPress."

// Avoid
"purpose": "Blog post about WordPress."

2. Use abstract as the LLM executive summary

abstract should match your meta description. It is the first field AI systems should read. Keep it factual, under 160 characters when possible.

3. Author editorial keyFacts

Write 3–7 atomic, generalizable facts. Each should stand alone as a citable statement.

// Good
"GEO optimizes content for AI citation, not just traditional search rankings."

// Avoid (raw H2 text)
"When users ask: What is the best AI visibility platform?"

4. Classify claims by evidence type

• internal — evidence on your own site (features page, docs, this article)
• external — third-party sources (Schema.org, llmstxt.org, research papers)

5. Keep contentFreshness accurate

Update dateModified when facts change. Stale manifests reduce AI citation confidence.

6. Link related pages

Use relatedPages to build a topic graph. Prefer extends and tutorial relationships over generic links.

7. Declare discovery assets

Point to site-level files (entity.json, llms.txt) in the assets block so AI systems can traverse your discovery layer.

8. Set citation policy

Use citationPolicy to tell AI systems how to handle conflicts — e.g. prefer entity.json over marketing copy.