The method behind every number.
Search results change with place, question, time and platform. Our method separates what we observed from what we inferred, so you can judge the findings for yourself.
1. AI measurement
We ask realistic buyer questions on the assistants named in your scope, repeating each question across runs and paraphrases, and keep every response. We separate four outcomes: mentioned, cited, recommended, and correctly described. Each rate is reported with its numerator and denominator (for example, named in 42 of 100 responses) and a confidence interval, which expresses sampling uncertainty under stated assumptions. Intervals do not predict what any single future answer will say. Assistants differ in how they retrieve information and how well location can be controlled, so we do not claim every assistant is queried identically. The exact interval method, run counts and collection window are stated in each report.
2. GEO Readiness
GEO means Generative Engine Optimization. The GEO Readiness Score is a rubric of changeable factors: crawler accessibility, page structure and usefulness, business identity, profile quality, consistency across sources, review-related data and supporting evidence. Bands: 0–39 Critical, 40–59 Weak, 60–79 Competitive, 80–100 Strong.
Readiness weights are heuristic priorities, not measured effects on search or AI rankings.
The weights come from our own judgment. They are not derived from a measured effect on any search or AI system, and a high readiness score does not predict higher visibility.
3. Local map check (supporting)
This supporting check adds local context to the AI findings and is reported separately. We search Google Maps from a sampled grid of points across your service area, using the buyer queries named in your scope, and record whether and where your business appears. Where included, we compare open-hours and closed-hours states. Points we could not measure are marked as not measured, never counted as misses. Distance from a searcher can change local results, which is why we measure many locations rather than one. Share of Local Voice and average rank figures appear only when the report actually calculates them.
4. Evidence tiers
- Source observation: something saved and reviewable, such as a page, a listing, or an AI response.
- Reproducible technical check: a test anyone can rerun, such as whether a page is crawlable.
- Expert judgment: our interpretation, labeled as such.
- Experimental idea: a tactic with weak evidence, listed lower or omitted.
5. Competitors
Competitors are chosen with you. We measure them under the same conditions where feasible. We do not know their traffic, revenue or private data.
6. Change over time
A later report can be compared only where the sample, questions, platforms and grid are compatible. Platforms change without notice, so some movement is drift rather than the effect of anything you did. A change in a measure is not proof that a particular fix caused it.
7. Data handling
How we retain, redact and restrict access to audit data is described in your service agreement and our privacy notice.
What we do not claim
- We do not know Google's or any assistant's ranking weights.
- Structured data or an llms.txt file does not guarantee AI citations; we check machine-readable business information as one factor among others.
- Some ranking influences are not visible to third parties.