Obility Editorial · · 4 min read
Begin with a decision
A useful measurement program starts with a question your team can act on. “Are we visible in AI search?” is too broad to guide a content review. “Do answers about our category explain the use case on this page?” gives you something concrete to investigate.
Write down the audience, product and buying stage before collecting answers. Keep the initial scope small enough that someone can read the observations rather than only look at a score.
Choose prompts from customer questions
Start with questions from sales conversations, support requests and your own content planning. Separate category discovery, comparison and implementation questions. They serve different needs and should not be collapsed into one ranking.
Preserve the original wording. If you rewrite a question, record the change. Otherwise a change in the prompt can look like a change in your brand’s visibility.
Save the evidence behind the observation
Keep the prompt, engine, date, relevant context, answer text and cited URLs together. Note whether you used an account and what location or language you selected. Treat each answer as an observation under those conditions.
Record a brand mention separately from a recommendation or a citation to your website. Those are different outcomes. An answer may mention your name without recommending you, or cite your documentation without naming your product.
Compare before taking action
Repeat observations on a consistent schedule and look for patterns worth investigating. Do not treat one favorable answer as proof of improvement or one missing mention as proof of a persistent problem.
When you make a content change, record which page changed, why and when. Compare subsequent observations without assuming your change caused every movement. Other sources and the answer engine itself may have changed too.
Turn the baseline into a work queue
For each recurring gap, assign a next action: review an existing page, clarify a product fact, investigate a cited source or collect more evidence. Give the action an owner and a follow-up date.
The result should be a manageable set of decisions, not a longer spreadsheet nobody reads. A baseline is useful when it changes what the team does next.