Vantage is a count, not a model. For each industry, a fixed set of buyer questions is put to each engine every week, every eligible company is checked against every answer, and the score is the share of answers that named the company. Everything below says exactly what that means and where it stops.
Methodology version pending publication. Contract version 2026-10-23. Measurement is run by Answyn; publication and presentation are by Tilio.
The terms used on every page, table, chart and email, defined once.
Which engines can be run, how a week is scanned, and when a week is published.
An industry runs a stated set of engines, listed on its page with the answers each contributed. The engines that can be run are ChatGPT, Google AI Overviews, Google AI Mode, Google Gemini, Perplexity, Microsoft Copilot, Anthropic Claude. Claude is collected weekly rather than daily, which suits a weekly scan.
One scan per industry per week, Monday to Sunday UTC, in the industry’s market. Every question is put to every engine in that window. One scan serves every company in the industry; no company is measured on a different set of answers from its competitors.
A week is published only if the scan completed at least 95% of the attempts it set out to make and at least 90% on every engine. A week that falls short is held; the previous published week stays the headline and the page says so. Published weeks are immutable: a figure does not change after it is published.
Question sets, company lists and alias lists are versioned. A change to any of them starts a new comparison period: trends show a break, the first week after the change shows no week-on-week movement, and the page states which version was used.
What a figure can and cannot tell you.
Tilio’s wider measurement method, for the tracked prompts it runs for clients, is described in How we measure AI visibility. Vantage follows the same standards: stated denominators, missing answers identified rather than counted, and sample changes declared.
No. Visibility counts answers that named the company. A mention can be a recommendation, a comparison or a caveat. Where an answer is classified as a recommendation that is shown separately; where no classification exists, nothing is claimed.
Each engine answers in its own way and each has its own denominator. Google AI Overviews, for example, often shows no AI answer for a question; those searches are reported and left out of its denominator rather than counted as an absence.
No. Each industry's eligible company list is fixed before the scan, every company is measured against the same answers, and nobody can pay to be included, excluded or moved. A listing does not mean the company is a Tilio client.
Changes are shown only between weeks measured with the same question set and the same company list. A new question-set version, a cohort change or a first published week shows no change, and says so.
No. 0.0% means the company is on the eligible list and was named in none of the eligible answers: a real result. A company that is not on the list has not been measured at all, and can request inclusion.
