Separate search access from model training
OpenAI documents OAI-SearchBot for search discovery and GPTBot for content that may be used in model training. Their robots settings are independent. A policy choice about training should not be confused with a search visibility setting.
Check accessibility first
Review returned status codes, robots rules, public HTML, canonical URLs, and internal links. Access is a necessary practical check, not a guarantee that a platform will select the page. Private or failed pages should not be assumed to be available because the homepage loads.
Make the page useful
State the real offer, provider, customer, and relevant conditions. Answer the visitor's task with supporting evidence. A service page might explain deliverables, access requirements, and scope; a case study should define the outcome and measurement period.
Keep the language grounded
There is no public formula in the cited documentation that lets a consultant compute a page's citation probability from schema density, paragraph length, or Wikidata links. Do not present a proprietary-weighting claim as established fact.
An observation log
| Record | Example of the information needed |
|---|---|
| Prompt | Exact question and follow-up context |
| Environment | Date, country, account context, and search mode |
| Source | Cited URL and the claim it supports |
| Outcome | Referral visit, inquiry, or no observed action |
Evaluate improvements responsibly
Record changes, repeat a defined set of observations, and compare like-for-like results. If several changes ship together, avoid claiming one caused all improvements. The practical goal is accurate discovery and useful business outcomes, not a simulated score.