← BlogGuide · · 10 min

The Check: a complete guide to the AI Buyer Audit

What The Check is, what it examines, how it works, what you receive and what happens after. The plain guide to CheckMarque AI's AI Buyer Audit.

This guide explains what you buy when you buy The Check, CheckMarque AI's AI Buyer Audit: what it looks at, how it runs, what you receive and what happens afterwards. It is written for the person who has to decide whether it is worth doing, so it explains the method only as far as you need to judge the result. The full detail is on the methodology page. For why the shift matters, read the AI buyer journey; for the discipline behind the audit, the guide to AI buyer intelligence.

What is an AI Buyer Audit?

Buyers now ask AI to research, compare and shortlist before they contact a business. One conversation can decide who gets considered, so it is worth knowing what is said in it.

An AI Buyer Audit reads that conversation from your side. It answers three questions about your business, for the buyers you want to win:

  • Found: when a buyer describes their need without naming you, does AI name you?
  • Understood: when AI names you, does it describe you correctly, and show how you fit this buyer: their problem, their circumstances and the questions they need answered?
  • Actionable: when the buyer is ready, can they act on that answer, with you?

Then it shows why, and what to change.

The Check is our audit. It is the entry point to everything else we do, because monitoring and content both build on its findings and its prompt set.

Why it matters to know how AI represents you

These conversations happen out of view, and the usual tools do not show them. Search rankings measure a list of links; AI writes one answer. A page can rank first in search and still be left out of the answer, or described wrongly in it.

Each gate is worth something different, and each can go wrong in its own way:

GateThe question it answersWhen it worksWhen it does not
FoundAre we even considered?A place on the shortlist, where the decision now starts.Buyers shortlist without you.
UnderstoodAre we described as we are, for this buyer?Buyers arrive understanding how you fit their situation, with their questions already answered.Wrong facts, a weaker price position, or being framed as a poorer fit for their situation. This can cost more than absence, because you fall out of the buyer's AI journey before they reach you.
ActionableWhen a buyer is ready, do they reach us?The ready buyer lands where they can act, with you.The buyer is ready but lands with a rival, on a platform that keeps the margin, or at a form their agent cannot complete.

For considered, higher-value purchases, one buyer won back can pay for the audit. A wrong answer fixed at its source is fixed once, not in every sales conversation that follows.

What The Check examines

The gates ask what is true about you in AI answers. These five areas are where we look to answer them. Evidence from one area usually speaks to more than one gate.

  1. Visibility and prompt coverage. It starts from a map of what your buyers ask AI and what you currently cover. Prompts drawn from that map are run as full conversations on ChatGPT, Gemini, Claude and Perplexity, recording whether you surface, where, and how you are described, and which other companies surface instead.
  2. Identity and authority. Whether machines resolve you to the right company, and whether your domain is treated as a source worth drawing from: entity presence, structured data, naming consistency, and whether your site is cited in answers about your category.
  3. AI crawler access and readability. Whether each major AI crawler can reach the pages that matter, whether your robots.txt and hosting settings let them in, and whether the facts buyers need are in plain text rather than PDFs, images or forms.
  4. Agent readiness. Whether a buyer's agent could act with you today: find the right page, read prices, terms and availability, and get past forms and bot protection to quote, book or order.
  5. Sources and evidence. Which of your pages and which third-party sources each engine draws on, which facts it states correctly or wrongly, and where the wrong ones come from.

The same five areas are run against your competitors. That is what produces the competitor decomposition.

The three gates, in more detail

Every answer an engine gives about you is read at the three gates, and each gate is scored on its own. There is no single blended score, because a blend would hide which gate matters most for you.

Three gates and their statesEach gate is read on its own
Gate 1Found

When a buyer describes the problem without your name, does AI name you?

  • CreditedNamed, with one of your own pages as the source
  • UnsourcedNamed, but from memory or other people's pages
  • Ghost-citedYour page is used, your name never appears
  • OverlookedNot there at all, or the problem is solved another way (framed out)
Gate 2Understood

When AI names you, does it get you right?

  • MatchesAccurate and fairly framed
  • MisrepresentedWrong, such as an old fee from a stale listing
  • UndersoldA fact the buyer needs is missing
  • Out-positionedAccurate, but a rival is set up as the better choice
Gate 3Actionable

When the buyer is ready to act, where does the answer send them?

  • To youRouted to a page where they can book, or an agent books for them
  • Stuck at your doorNamed, but with no way to act
  • LostSent to a rival or a comparison site, dropped at the last question, or never considered
PassesWeakFails

Behind the gates sit 16 states and 26 measures, and a cause chain that links each change to what it should move. You do not need them to read your report, but they are why every finding can say what to change and what it should move. They are set out on the methodology page.

How your buyers and their prompts are defined

The Check does not start from a fixed list of prompts. It starts from your business.

  • Discovery, before we ask you for anything. From your URL, we read your site and work out your company, industry, offer and products, and a first set of buyer prompts. This is also where we see the buyer AI already imagines for you, which is the buyer it recommends you to.
  • Your buyers, with you. You correct what we found and add what a website cannot tell us: what your buyers actually ask AI, your real competitors, the pages that win deals, and what you are trying to achieve.

From this we build the prompt map. It decides every prompt put to the engines, which findings rank as priorities, which competitors the decomposition runs against, and which business goal the insights are measured against.

How the AI engines are assessed

Each prompt is sent as written to every major engine, including ChatGPT, Gemini, Claude and Perplexity. We record whether you appear, where, how you are described, which competitors appear alongside you, and which sources each engine cites.

Each prompt is run several times on each engine, because answers vary from one run to the next. One answer from one engine on one day is a sample, not a finding. Asking repeatedly is what lets us tell a pattern from chance.

Engines are read separately, not averaged. Each has its own habits: some search before they answer, some answer from memory, and they cite different kinds of sources. A finding that holds on one engine and not on another is itself useful to know.

How evidence and sources are examined

Every claim in The Check carries its evidence: the engine that answered, the date, the prompt and the sources it drew on. That record is what lets you check any finding yourself.

Beyond what engines say, we look at why they say it:

  • which of your pages, and which third-party pages, the engines draw on;
  • which facts about you they state correctly, which they get wrong, and where the wrong ones come from;
  • which pages engines use without crediting you;
  • which facts a buyer needs that engines cannot find, for example because they sit in a PDF or behind a form.

You cannot edit an AI's answer. You can change what is behind it: your site, the sources engines read, and the facts a buyer needs. That is why the evidence matters more than the score.

How competitive positioning is evaluated

The competitor decomposition runs the same assessment against two sets of companies: the competitors you name, and the competitors the engines actually return. These are frequently different.

It shows where you stand against each one, who AI groups you with (often not the list you would choose), which unexpected rivals take your place, and what earns them their citations. It also shows open ground: buyer prompts nobody answers well yet, where a business can win fastest.

How findings are diagnosed

Individual signals describe symptoms. A finding explains the cause.

For each problem, the diagnosis connects what the engines said to what is behind it: a page that does not state the fact, a stale listing that does, a competitor with better proof on the point the buyer cares about, a crawler that is blocked. Causes can cross gates. A weak presence in the sources engines cite, for example, can keep you out of the answer and also make a rival look stronger when you are named.

Where the evidence is unstable, for example when engines disagree with each other or with themselves from run to run, we say so rather than call it.

What you receive

Six things, each traceable to the evidence behind it, and two free agentic checks while we are in early access.

  1. Your buyer and prompt map. Who AI thinks your buyers are, confirmed with you, and the prompts they actually give AI, from the first question to the follow-ups. If you want other buyers covered, we add them.
  2. The signal report. Every signal in our framework, assessed and scored for you, with the engine, date and source behind each one.
  3. Competitor decomposition. Where you stand against the competitors you named and the ones the engines return, and what earns them their place.
  4. Actionable insights. The causes that connect the findings, which of them matter for your business and your goals, and what each fix should move.
  5. A sequenced action plan. Every finding ordered by impact over effort: what to fix now, what to restructure next, and what to build over the quarter. Each item names the team that owns it.
  6. A walkthrough. We take your team through every finding: what it means, why it matters in your category, and where a competitor is ahead and on what. Questions are answered against the evidence.

Free while we are in early access:

  • Agent trial run. A simulated buyer agent tries to get a quote, book or buy with you, the way a real one would, and records exactly where it gets stuck.
  • Agent-readable facts check. Your key facts, such as prices, terms, availability and proof, checked against the formats agents read, with what is missing and where.

The Check starts at €899 and is priced on scope: your offer, your markets and the buyers you want to win.

How recommendations are prioritised

Findings are not labelled high, medium or low. Each one names the gate it blocks and is ranked by impact over effort, which is an order, not a score.

Findings that block everything else, such as an AI crawler shut out of your key pages, are marked and sequenced first. Many findings need only a small change. A team that does the few things that matter first moves further than one working through a long list.

Each finding is written as a task: what to change, who owns it, roughly how much effort it takes, and what it should move. Marketing gets content aimed at a specific finding, web and engineering get technical fixes, and sales learns what buyers will arrive believing.

What happens after the audit

After the walkthrough, you decide what runs on. We work in three steps.

StepWhat it isWho does it
AssessThe Check: where you stand, why, and what to change.Us, with your input
ImproveThe fixes in the action plan. Where a finding calls for new or better pages, content refresh: content we write with your subject matter experts, each piece aimed at the signal it should move.Your team, your agency or our industry partners for fixes; us for content refresh
MonitorPrompt monitoring: your prompt set re-run on the cadence you choose, each run compared with the one before.Us

Implementation of technical fixes is not part of The Check itself. Your team, your agency or our industry partners make them, with the action plan and the evidence in hand. Content refresh and prompt monitoring are separate deliverables you can add.

Ongoing monitoring and optimisation

None of this stays still. Engines update, the sources they read shift, buyers change how they ask, and rules change. Your business moves too, with new products, new markets and new rivals. A single audit is a photo of a moving picture.

Prompt monitoring turns the photo into a film. The same prompts are re-run on your cadence and compared like with like, so you see what moved, what did not, and the next thing worth fixing. Every fix in the action plan names what it should move, so the next run either confirms it worked or says the finding needs a different answer.

You can extend the prompt set at any time, as you launch products, enter markets or find new buyer prompts worth tracking. Under a retainer you can combine all three deliverables, each on its own schedule: monitoring fortnightly with a full audit quarterly, for example, and content refresh only when there is something to move.

Where to start

The Check starts with a free demo. Send us your URL, and a buyer or prompt you care about if you have one. We show you the buyer AI imagines for your business and that buyer's conversation with AI: whether you were named, how you were described, and who was named instead. If it is worth going further, we scope The Check around the buyers and prompts that matter to you. If it is not, we say so.

Request a demo

Where to start

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.

CheckMarque AI

© 2026 CheckMarque AI. All rights reserved.PrivacyCookiesTermsLinkedIn