The prompts, what to record and how to read the result
The check takes about thirty minutes and needs no tool or subscription. The goal is not a single snapshot but a measurement you repeat every month with the same prompts, because the answers of AI tools change from day to day.
Prompts are written the way a buyer asks, not the way a company presents itself. One prompt per service, without your company name.
ChatGPT, Google Gemini, Perplexity and Google (AI Overviews). The same prompt text in each, with no extra explanation.
| Prompt | Tool | Mentions us (Y/N) | Which companies are named | Sources cited | Is the detail correct |
|---|---|---|---|---|---|
This is not a ranking problem but a sourcing one: there is not enough confirmation about you in the places models read. The first steps are a page that states plainly what you do and for whom, and the same data in directories and trade portals.
The model picked up an old or someone else's detail. It is corrected by consistency: the same name, the same activity, the same area of work everywhere the company appears.
A sign the site is not written in a form a machine can cite. Answers to concrete questions, structured data and llms.txt all help.
| Month | Prompts we appear in | Out of | Note |
|---|---|---|---|