What we measure. How we score it.
We call it The Check: an interactive report with a walkthrough, the AI Buyer Audit run by Marque. Every finding carries its evidence, the change to make, who owns it and roughly how much effort it takes.
Three gates, one question each
Each gate is one question with one pass rule, judged on its own. A business can pass one gate and fail another for the same buyer.
Does AI name you?
You are in the room for every one of those questions.
When a buyer describes the problem without naming you, are you Mentioned, Shortlisted or Recommended? Or Absent, framed out by another kind of solution, or used as a source without your name?
Does it get you right?
The answer sells you accurately when you are not there.
Facts and framing, judged against what you confirmed is true and what this buyer needs. Accurate, complete, and not set up as the weaker choice. This is the one companies underestimate.
Can the buyer act on it, with you?
A ready buyer, or their AI agent, lands where they can act.
When a buyer is ready, does the answer send them to a page where they can book, quote or buy? Or does an AI agent complete it for them? Anything else is a ready buyer lost.
Buyers enter anywhere.
Some describe a need and meet Gate 1. Others ask for you by name and start at Gate 2. Red marks where they are lost or misinformed.
From the first question to the final choice.
If your buyers go online at any point while they decide, three scores show how AI carries them forward, and where your next effort counts most.
Each gate gets its own score: the share of counted answers that pass it.
Each gate shows a different part of how AI treats you, and each is owned by a different team. Scored separately, they show where buyers move forward and where they stop, so you know where to focus your effort and strategy, and the first fix goes where it moves the most. Three separate scores keep that picture clear.
Each square is one counted answer, coloured by its state: green passes, amber and red do not.
PassesCredited or Unsourced
PassesMatches
PassesRouted to you or Acted for you
Twenty-six diagnostic measures are scored today, with more in beta, and the set is agreed with you around what you need to know. Each is tied to the gate it explains, so a gate score always opens into the scores behind it.
3 gates, 16 states, 2 ways in.
Each gate lists its states from best to worst. Gate 3 is the outcome: every decision point ends in one of its eight states. It records what the answer did, not what the buyer did next.
How does the buyer start?
Gate 2 needs a name.
You can only be described if you are named or asked about. If you are not named, there is no Gate 2 state, and at a decision point the outcome is Not considered.
Gate 3 applies at every decision point.
Whatever happened at the first two. A buyer can reach you after being misrepresented or undersold; they arrive misinformed. A named buyer can end in any outcome except Not considered.
Every gate score opens into the measures behind it.
Twenty-six measures are scored today, with more in beta, and we agree the set with you around what you need to know. Each is tied to one gate and kept only because it points to a change. Pick one to see what it means and what feeds it.
Citation share
ScoresNamed answers that cite your own pages
What it tells youWhether you control what is said about you (Credited vs Unsourced)
Fix a cause, its effects move, the gate score follows.
Each gate score is made of effects: what the engines say and do, measured in the captured answers. You cannot edit those directly. Behind them sit causes: what is on your site and in the sources engines read. Those you can change.
Every outcome also reads as a path, one line per gate
Sent to a rival
Found: Credited (shortlisted) · Understood: Undersold (never surfaced)
The framework is the third of six layers.
Each layer answers one question, in this order, and feeds the next. A cause only matters once it explains a gate you are failing.
Who is asking?
Buyer personas, buyer moments, confirmed factsWhat did the engines say?
Every captured answer, with engine, date, question and sourcesWhere do you stand, gate by gate?
A state at each gate, the outcome, three gate scoresWhy do you stand there?
Causes, traced to their sourceWhat should change, and who does it?
Findings: the change, its owner, its effortDid it move?
The same questions re-run and compared; the next thing to fixEvery change read through people and behaviour, technology and AI, and brand and narrative.
Being found and understood by AI is not a technical problem, a brand problem or a behavioural one. It is all three at once.
Model frameworks
What the systems can do. A new model relies more on its agentic capabilities: it browses, compares and acts for the buyer.
A new model version reads your pages and sources afresh and forms its own picture of you. Clear, consistent facts carry over best.
Each version retells your story. Where your positioning is stated plainly and the same everywhere, it comes through intact.
Buyers asking the same question get the new model's account of you, so it helps to know what that account says.
Send us your URL. We will take it from there.
Add a buyer or a prompt you care about, and we will build the demo around it. You see what AI tells your buyers before you decide anything.