Ask an AI assistant the question that would hire you, and count whose name comes back in the answer. That single test tells you more than a year of rank reports, and it takes ten minutes. This page walks through the full diagnosis, three tests from quickest to most conclusive, what each result means, and what to do when the answer turns out to be your competitor. It’s written as a companion to our guide on Microsoft Clarity and AI visibility, which covers the tool side of this in depth.

The Three Tests That Answer the Question

You can settle this question with three checks, ordered here from fastest to most conclusive.

  • The prompt test: Ask the assistants directly, the way a customer would, and record who gets named.

  • The referral check: Look at your analytics for traffic arriving from AI platforms, because a recommendation always leaves a trail.

  • The retrieval check: Inspect what the assistants actually pull from your site, through Microsoft Clarity or server logs.

Run all three and you’ll know exactly where you stand, including whether a competitor owns your category’s answers right now.

Test One: Ask the Assistant the Way a Customer Would

The prompt test is a manual audit, and it works because it reproduces the exact moment a customer chooses.

  • Open a fresh chat in ChatGPT, Claude, Perplexity, and Gemini, and type the question a customer with a broken pipe, a traffic ticket, or a leaking roof would type.

  • Record which businesses get named, which get described warmly, and which get dismissed with a hedge.

  • Repeat across several days, because AI answers shift with each retrieval.

Test Two: Check Your Analytics for AI Referrals

A recommendation that turns into a customer always shows up in your analytics as referral traffic from an AI platform.

  • Look for chatgpt.com, perplexity.ai, copilot.microsoft.com, and gemini.google.com in your referral sources.

  • A steady stream of these visits means assistants are already naming you somewhere upstream, before the click.

  • Zero AI referrals in a category where competitors are seeing them is the quiet alarm. The recommendation is happening. It is just pointing at them.

We watch these sources on our own domain first, because our AI Briefing content is produced with a custom in-house AI agent and we would not ask a client to track anything we ignore ourselves. For the tool that makes this check free, see our companion piece on what Microsoft Clarity shows.

Test Three: Read the Retrieval Layer with Microsoft Clarity

Referral traffic only shows the recommendations that became clicks. Microsoft Clarity’s AI Visibility layer shows the ones still forming.

  • It reports which AI systems are crawling your site, which of your pages get cited in generated answers, and how your share of answers compares to competitors.

  • It surfaces the actual questions the AI worked through to build each response, which are rarely the keywords owners think they rank for.

  • Two cautions from our own deployment. The setup has traps that quietly break your data, and Clarity reflects Microsoft’s AI surfaces, so it is one clear window rather than the whole building.

Before Clarity went on any client site, it went on ours. We don’t recommend anything we haven’t run ourselves, and an agency like ours built on a finance background would rather measure than guess.

The Verification Nobody Can Argue With

Server logs are the only source that counts every AI platform at once, checked against the platforms’ own official IP lists.

  • We pulled 30 consecutive days of logs for our client Weiland Button, PLC, 431,753 total requests, and isolated verified AI retrieval.

  • The firm’s content was retrieved 2,806 times in a single month, over 97% of it matching the official address lists OpenAI and Perplexity publish.

  • The steady state has held at 560 to 640 citations per week across 24 Virginia jurisdictions.

Most agencies have never pulled this data for any client, which is why most agencies can’t tell you whether AI is recommending you or your competitor. We publish the method and the numbers because the answer should come from measurement, not from a vendor’s promise.

What Each Result Means

The three outcomes carry very different news, and each one changes what you do next.

  • AI names you: You hold an endorsement, and endorsements compound. Assistants return to sources that answer well, so every month of citations deepens the moat.

  • AI names a competitor: Their pages currently answer the questions better than yours, in the way the machines read. Every month untreated, their advantage compounds the same way.

  • AI names nobody: Congratulations, the shelf is empty. The first business with the right answer owns the category, which is exactly how Weiland Button’s pages for Nottoway, Stafford, Sussex, and Dinwiddie counties became the cited answer in a near-empty field.

The worst position is not losing the recommendation. It’s never checking, while the decision moves upstream and your phone stays quiet for reasons no traditional report will ever explain.

If You Need AI Expertise, Lift Conversions is Here

Lift Conversions has run month-to-month with no contracts since 2014, so a client can tell by month four whether the work is moving their share of AI answers, and more than 15% of our current clients have stayed past ten years on exactly those terms.

Published On: October 7th, 2026 / Categories: AI Visibility & AEO / Tags: , , , /

Subscribe to our Marketing Newsletter.

What changed in search and how to adapt now, in your inbox weekly.

We respect your inbox. Unsubscribe anytime. See our Privacy Policy.