Building a visibility baseline a team can compare over time.
An illustrative study of a marketing team defining prompt cohorts, market context, and reporting rules before expanding AI search measurement.
This is a fictional worked example. Any figures are synthetic demonstration data, not achieved client results. The scope, timeline, and interpretation explain how a real engagement could be documented.
“Are we becoming easier to find in the markets we serve?”
The challenge
Imagine a marketing team comparing screenshots gathered by different people, in different markets, using different questions. The answers are interesting, but an apparent improvement could come from a changed prompt or a personalized session. The team needs a consistent basis for deciding what to do next.
A proposed approach
Define the observation
Choose a buyer segment and market for each cohort. Record the question, language, engine surface, date, available model information, and session conditions. Separate branded checks from unbranded discovery and comparison questions.
Preserve the sample
Store the observed answer and its linked sources alongside the sampling details. Count brand mentions, owned citations, and factual errors separately. Use a stable core of questions and label any new exploratory cohort.
Connect findings to delivery
Review recurring issues and identify the pages or facts that can be improved. Assign owners, record publication dates, and annotate reporting changes. Discuss referrals and qualified inquiries in their own sections.
How we would evaluate the work
- Compare like-for-like cohorts before combining results.
- Show answer counts alongside every rate.
- Record missing observations and sampling changes.
- Treat changes as evidence to investigate, without claiming a single edit caused them.
If the sample changes materially, establish a new baseline and keep the earlier series visible. A transparent discontinuity is more useful than a smooth chart that hides a different method.
This educational example was created by Citevia. For our measurement principles, read how to measure AI search visibility.
