What we measure, and how it's read.
We measure whether your brand is surfaced across the specific moments a real buyer asks: a direct search, an open-ended recommendation, a specific-need question and a real-time availability check.
Preference is measured per model because a brand can be the default answer in one model and an afterthought in another.
We test the same buying question repeatedly and across phrasings, so the index reflects stable preference rather than a single lucky generation.
Competitors are scored on the same scale so you can see how your brand's MPI score compares model by model.
Stage: Quantify.
Machine Preference Index sits in the Quantify stage alongside Share of Model — it's the strength-of-preference layer behind the raw mention count.
See the full QUEEN framework →You're losing revenue to AI you can't see.
A mention that arrives third, hedged and uncited rarely converts. Preference — not presence — is what decides which brand an AI model puts forward when a buyer asks for one recommendation.
Tracking preference tells you whether your visibility work is actually moving you towards being chosen, or just towards being listed — the exact gap that let a regional QSR chain out-rank a global brand in grounded answers.
Machine Preference Index, answered.
↳ What is the Machine Preference Index?
The Machine Preference Index measures how often your brand surfaces across AI models — accounting for position, framing and confidence, not just whether your name appears.
↳ How is Machine Preference Index different from Share of Model?
Share of Model counts how often you are recommended. Machine Preference Index measures how often your brand surfaces in the answers that lead to those recommendations.
↳ Why does AI prefer my competitor over me?
Usually because the model can retrieve cleaner, more structured signals about them, because third-party sources corroborate their claims more consistently, or because their public content answers the buying question in the form the model retrieves. The index quantifies the gap; the Uncover stage explains it.
↳ Can Machine Preference Index be improved?
Yes. Preference responds to clearer product facts, structured data AI models can trust, corroborating third-party sources and content that answers buying questions directly. MARK prioritises those fixes by expected impact on the index.
Get your AI preference benchmarked — free.
Get your AI preference benchmarked — freeHow often your brand is recommended versus competitors and interceptors in your category, refreshed monthly.
How ready your brand is to be selected by autonomous AI agents, across both AEO and ACO.
How AI shopping agents actually experience your site — and a competitor's — and where it breaks.
What AI models actually believe about your brand — not just whether they recommend it.
Continuous monitoring of the data and content feeds AI models pull from, so drift never silently breaks visibility.