How we measure AI visibility
Tilio checks whether your business appears when AI answers questions relevant to your buyers. We examine the descriptions, recommendations and sources behind those answers, then use the findings to prioritise work.
Scope
Start with buyer questions
Measurement starts with an agreed scope: the questions, competitors, markets and engines being monitored. A result only makes sense within that scope.
Your prompt universe is the broader collection of questions buyers could ask about your category. It includes discovering suppliers, comparing options, checking prices and deciding whether a provider suits their needs.
The tracked set is a sample drawn from that universe. We group questions by commercial theme and intent so that an important service or buying decision does not disappear inside one overall score.
Prompt planning starts with your offer and customers. Sales conversations, common objections and questions about suitability can help identify useful themes. A proposed prompt remains a research hypothesis; including it in a tracker does not establish how often customers ask it.
The right sample size depends on your services, markets and scope. More variations of the same question do not automatically produce better coverage. We need enough breadth to support decisions and enough consistency to compare results over time.
Prompt types
Separate discovery from brand checks
Unbranded questions do not name your business. They test whether an answer brings you into the conversation, such as asking for suppliers of a particular service.
Branded questions name you or a competitor. They help examine how the answer describes a known business, explains its offer or compares it with another option.
These answer different questions. Appearing when someone explicitly asks about your business is different from appearing in a shortlist without that cue. Reports should keep those views separate when the distinction changes the interpretation.
Measure different signals
Several signals can appear in the same answer, but they mean different things:
| Signal | What it tells us |
|---|---|
| Brand mention | The answer names your business. |
| Recommendation | The answer presents your business as an option for the buyer's stated need. |
| Citation | The answer links to a source. We distinguish links to your website from links to third-party coverage. |
| Description | How the answer explains your services, suitability or other business facts. |
| Competitor presence | Which alternatives appear for the same question and how they are presented. |
A link to your website is not automatically an endorsement. A recommendation can also appear without a link. Those distinctions matter when deciding what to improve.
Interpretation
Compare fairly over time
A percentage needs a clear basis: which questions and engines were included, the period covered and how many answers were available for analysis. Missing responses should be identified rather than silently treated as brand absence.
Comparisons are most useful when brands are assessed against the same questions and collection conditions. Changes to the sample, coverage or classification can affect a reported trend, so they need to be considered alongside the movement.
One answer is an observation. Repeated observations can help identify patterns, but still represent a sample of what people might see. Differences between engines or topics can be more useful than a combined headline figure.
Next actions
Turn findings into priorities
The next action depends on the gap. An inaccurate explanation may point to unclear business information. Weak coverage of a buying question may call for a better commercial page. A relevant third-party source may suggest an outreach opportunity.
These are priorities to investigate, not proof that one missing page caused poor visibility. Gain, our monthly AEO programme, connects this analysis to the work included in your scope.
Referral visits and enquiries provide a separate view of business impact where tracking is available. A citation alone does not establish a visit, a lead or revenue. Our AI search attribution setup puts that tracking in place. See how working with Tilio operates for reporting and delivery.
FAQs
Does the score measure our whole market?
No. It describes the monitored sample and period, rather than every buyer or AI answer.
Can a rise prove that your work caused it?
Not by itself. Other website changes, competitors and changes within AI systems can affect results. We review those factors before drawing conclusions.
Which engines will you track?
We agree coverage for your programme. Results should identify the engines being measured so you can interpret them properly.
Can you guarantee recommendations?
No. Measurement helps us make better decisions; it does not give us control over AI answers.