Key takeaways
- Markup must match the page.
- Search and training controls differ.
- Test production access.
Check readable responses
An important URL should return useful text and working links without login. Inspect the actual production response. A local preview alone cannot prove that a hosting firewall or crawler rule allows access.
Describe the visible page
Use appropriate types such as Organization, Article or BreadcrumbList. Include accurate URLs and relationships. Do not insert facts that visitors cannot see. Valid markup is not a guarantee of a search feature.
Treat crawler controls separately
OpenAI documents search-related and training-related crawlers separately. Perplexity publishes its own crawler guidance. Read the current provider documentation before changing robots.txt or firewall rules; one agent does not cover every experience.
Avoid machine-only shortcuts
Do not hide keyword blocks or fabricate reviews. Google says special AI markup is not required for its AI search features. A machine-readable file cannot replace accurate, accessible information for readers.
Test detail pages too
Inspect robots.txt, HTTP status, canonical URL, visible content and links. Use Search Console and relevant structured-data tools after deployment. Nested routes can fail even when the homepage works.
References and further reading
Primary documentation and product references. Practical checklists and worked examples above are editorial guidance, not claims about private ranking algorithms.