Field method · Not experiment results
Observe AI discovery without inventing a ranking.
This is our proposed observation protocol. No prompts have been run or performance rates claimed in this article.
Use this method within a marketing workflow. The SEO / GEO capability notes explain where search visibility fits.
1. Fix the question and context
Choose an actual buyer question, such as “What information should I prepare for a bilingual manufacturer website project?” Record whether the brand was named in the prompt. Keep market, language, signed-in state, memory/search settings and new versus continued conversation consistent. Never compare a brand-named prompt with an unprompted discovery question as if they were the same test.
2. Use an empty observation sheet
| Prompt text | Date and timezone | Platform | Model if available | Language | Context | Brand mentioned | Site cited | Cited URL | Notes |
|---|---|---|---|---|---|---|---|---|---|
| — | — | — | — | — | — | — | — | — | — |
Blank template, not observed data. Use yes/no/unclear for mentions and citations; preserve the response or screenshot privately and link only publishable evidence. Unknown model information stays unknown.
Download blank CSV ↓3. Repeat and interpret cautiously
Record a baseline before a scoped page change. Repeat the same question under recorded conditions after the change. Choose the schedule in advance and retain negative observations, not just favorable answers. Report the observed sample and its limitations; model and index changes can be alternative explanations.
A mention is not a citation. A citation is not a visit. A visit is not a qualified enquiry. Link each step only when its own analytics or actual record supports it, with appropriate privacy controls.
Official guidance versus our method
Google states that existing SEO practices apply to its AI search features and that no special AI markup is required. OpenAI distinguishes its search crawler from its training crawler. These are platform descriptions; our observation protocol is a team method and cannot guarantee inclusion.
Google Search Central: AI features · OpenAI crawler documentation
Sources checked 27 September 2026. No security or crawler policy changes are implied.