AI visibility audit

AI visibility audit: find out what ChatGPT, Perplexity and Gemini say about your business.

Before you spend a dollar on being found in AI search, find out where you stand. An AI visibility audit answers one question: when a customer asks an assistant for a business like yours, do you come up, and is what it says correct?

The short answer

An AI visibility audit checks what ChatGPT, Perplexity, Gemini and Claude say when asked for a business like yours; what your site looks like to a crawler without JavaScript; whether your llms.txt and schema exist and say the right thing; and whether your name, phone and address match across the web. We deliver it as Verdict, a written, sourced audit with a letter grade, and as ongoing sampling in ReputeLift.

What an AI visibility audit looks at

This is the check, not the fix. The fix is answer engine optimization. An audit is worth doing first because most businesses are guessing about where they stand, and the guess is usually wrong in one of two directions: either they assume the assistants have never heard of them, or they assume the assistants get their details right. A good audit looks at four things.

  1. What the assistants say. Ask ChatGPT, Perplexity, Gemini and Claude the questions a customer would type: "a good roofer in Provo", "who does laser hair removal in Sandy", "is [your business name] any good". Record whether you are named, who is named instead, and whether the phone number, address and services attributed to you are correct. Assistants vary their answers from day to day, so one run is a snapshot, not a verdict.
  2. What your site looks like to a crawler. Many assistants and crawlers fetch a page without running JavaScript. If your site is built so the headings and text only appear once a script runs, a crawler sees a title and an empty body. The audit fetches each key page twice, once in a real browser and once without JavaScript, and compares the title, the headings and the word count. A big gap means the assistant is reading a different page from the one your customers see.
  3. Whether your llms.txt and schema exist and say the right thing. Schema.org structured data tells a machine what your business is, where it is and what it offers; an llms.txt gives a language model a curated map of the site. The audit checks that both exist, that they agree with the visible page, and that the llms.txt does not leak anything it should not (internal URLs, staging sites, prices you have not published).
  4. Whether your name, phone and address match. The assistant is trying to decide whether the business on your site, the one on Google Business Profile and the one on Facebook are the same business. A different phone format, an old address or a trading name on one and a legal name on another all count against you.

What our audit delivers

Verdict: the written, sourced audit

Verdict is a written website audit with five scored categories and one letter grade. Its AI read section fetches your pages twice, in a real browser and as a crawler without JavaScript, and diffs the title, headings and word count. It checks whether you publish an llms.txt and whether it leaks anything. Every factual claim in the report carries a numbered, dated source you can open yourself, and if a meaningful claim is unsourced or wrong you get your money back. It includes a competitor read and recommendations sorted into quick wins, strategic and long term. Prices are published on the Verdict page; you buy it online and no call is required.

ReputeLift: the ongoing sampling

A single audit tells you where you stood on one day. ReputeLift's AI visibility check asks ChatGPT, Claude, Gemini and Perplexity a fixed set of questions a customer in your category would ask, and stores the answers with the date. Re-run on a schedule, it shows whether you have started appearing, stopped appearing, or are being named with the wrong details. It sits alongside ReputeLift's review requests, which ask every customer the same way after the job, with no gating and no sentiment filter, because Google reviews are one of the strongest signals an assistant reads for a local business.

How to read the result

Three things to look for, in order.

  • Named, not named, or named wrong. Not being named is the common case and it is fixable. Being named with a wrong phone number or a closed location is worse, because a customer acts on it. Fix wrong before you chase absent.
  • The crawler gap. If the no-JavaScript fetch shows a fraction of the words the browser shows, that is your first job. No amount of content work helps a page the assistant cannot read.
  • The consistency list. Every place your details disagree is a line item. Work through them; there is no shortcut.

Read a letter grade as a summary of the site, not of your business. A well-run company with a thin website gets a poor grade, and that is the point: the grade is telling you what the machines can see.

What an AI visibility audit cannot tell you

Verdict does not give backlink counts or keyword search volumes. Those need paid third-party data we do not pretend to have; where a recommendation would depend on them, the report says so and lists it as a limitation instead of guessing. It does not see your internal numbers: your close rate, your margins or your ad spend, so any revenue framing is a projection and is labelled as one. It holds prepared industry context for SaaS, local services and ecommerce only; outside those, the competitive read is thinner and the report says so.

No audit, ours or anyone's, can tell you why an assistant chose one business over another. The assistants do not publish their rules. What an audit can do is show you the inputs the assistant has to work with and where they are missing, broken or contradictory.

Do it yourself in 10 minutes

You can get a rough answer today without buying anything.

  1. Ask the assistants the same five questions. Open ChatGPT, Perplexity and Gemini. Ask each one: "a good [your trade] in [your city]", "who should I call for [your main service] near [your neighbourhood]", "is [your business name] any good", "what does [your business name] charge", and "[your business name] phone number". Write down who is named and whether your details are right. Do it again in a week; the answers will differ, and that is normal.
  2. Fetch your page with JavaScript off. In your browser's settings, turn off JavaScript and reload your homepage and your main service page. If the headings and text vanish or shrink to a sentence, a crawler is seeing the same thing.
  3. Check your llms.txt. Type your domain followed by /llms.txt into the address bar. Either you have one or you get a not-found page. If you have one, read it as a stranger would and check it does not name anything private. See what llms.txt is.
  4. Compare your phone number. Look at the number on your Google Business Profile, on your website footer and on your Facebook page. Same digits, same format, same area code? Do the same for your address and your business name.

If you find a crawler gap or a mismatched phone number, you have already found the two highest-payback fixes. If you want the written version with sources, a grade and the competitor read, that is Verdict. If you want it fixed and run, that is the answer engine optimization work, and it starts with a 15-minute conversation.

Find out where you stand before you spend anything

Buy a Verdict online for the written audit, or take fifteen minutes with Artie, the founder. Bring your business name and your city; you leave knowing whether the assistants name you and what they get wrong.

Book 15 minutes with Artie
Questions

Questions people ask

Does ChatGPT recommend my business?

The only way to know is to ask it the questions a customer would ask and record the answer. Ask for a business in your category and your city, ask about your business by name, and ask for your phone number. Repeat it in a week, because answers vary. Our audit does this across ChatGPT, Claude, Gemini and Perplexity and stores the answers with the date so you can see the trend.

How do I check if my business shows up in ChatGPT?

Ask it directly: "a good [your trade] in [your city]" and "is [your business name] any good". Then ask Perplexity and Gemini the same. If you are not named, check what your site looks like with JavaScript off and whether your name, phone and address match across Google Business Profile, your site and Facebook. Those are the usual reasons.

What is in an AI visibility report?

Ours (Verdict) contains a reliability header, five scored categories with one letter grade, an AI read that compares what a browser sees with what a crawler without JavaScript sees, an llms.txt check, a competitor read, and recommendations sorted into quick wins, strategic and long term. Every factual claim carries a numbered, dated source.

How much does an AI visibility audit cost?

Verdict has published prices on the Verdict page; you buy it online and no call is required. If a meaningful claim in the report is unsourced or wrong, you get your money back. Ongoing sampling in ReputeLift is scoped with the rest of the system on a 15-minute conversation.

Is an AI visibility audit the same as an SEO audit?

They overlap but they ask different questions. An SEO audit is about ranking pages in a list of results and often leans on backlink and keyword-volume data. An AI visibility audit is about whether an assistant can read your site, understand what your business is, and name it when asked. Verdict does not provide backlink counts or keyword volumes.

What can the audit not tell me?

It cannot tell you why an assistant chose a competitor, because the assistants do not publish their rules. Verdict does not see your internal numbers, does not give backlink counts or keyword volumes, and holds prepared industry context for SaaS, local services and ecommerce only. It shows you the inputs the machines have to work with and where they are missing or contradictory.

Related
15 minutes with Artie

Fifteen minutes with the founder. No sales rep in between.

Pick a time and it lands on Artie's calendar directly. You will get a confirmation by email and, if you leave a mobile number, a reminder by text. That booking system is the same one we run for clients.

  • Bring the thing that is slipping: missed calls, quotes nobody chased, reviews you meant to ask for.
  • Leave with a straight answer on whether it is worth automating, and what it would cost.
  • If it is not a fit, you will hear that too.

Loading the calendar. If it does not appear, open it here.