How to Audit Whether AI Actually Knows Your Business (A Founder's Field Guide)

Before a recent prospect became a client, I ran the standard pre-call diagnostic I always run. I typed every one of their priority keywords into four AI models to see how the company surfaced. They did not surface. Not once. Not for branded queries, not for category queries, not for the comparison searches their buyers were almost certainly running before contacting any vendor in their space. Their competitors were in the answers. They were not.

When I walked them through the findings on our first call, the founder's response was the one I get most often. He did not know AI visibility was something he could measure, and he had assumed his strong website and decent Google rankings meant he was covered. They did not, and he was not. That gap between what founders assume about their visibility and what is actually happening in AI answers is where most of the work sits right now.

I run a fractional CMO practice and I have done versions of this audit dozens of times across DTC, telehealth, and B2B services. What follows is the same diagnostic, stripped down so you can do it yourself in roughly an afternoon. You will not get a perfect SEO and AEO audit out of this, but you will know exactly where you stand, what is broken, and what to fix first. That is usually enough to stop guessing and start the actual work.

Start with the questions a real buyer would ask

Most founders audit AI visibility wrong on the first move. They type their own company name into ChatGPT, see something flattering or unflattering, and call that the answer. That is not the audit. That is vanity.

The real audit starts with the questions your buyer asks before they have ever heard of you. If you sell B2B compliance software to mid-market healthcare companies, your buyer is not typing your name. She is typing "best HIPAA compliance tools for mid-sized clinics" or "alternatives to [name of the incumbent everyone knows]." If you run a fractional CMO practice like mine, the prospect is asking "how do I know if I need a fractional CMO" or "fractional CMO versus marketing agency for a Series A company."

Write down ten of these questions. Not five. Ten. The first three will be obvious and the last seven are where the actual intelligence lives, because they force you to think about adjacent searches, comparison searches, and the searches people run when they are quietly checking whether you are real.

Then run every one of those ten queries through four models. I use ChatGPT, Claude, Gemini, and Perplexity. Each one indexes the web slightly differently and weights sources slightly differently, and the variance between them tells you something. If you appear in Perplexity but not in ChatGPT, that is a signal about which sources each model trusts. If you appear in none of them, that is a different and more urgent signal.

Save the answers. Screenshot them, paste them into a document, do whatever works for you, but capture them. You will need a baseline to measure against in ninety days.

Read the citations, not the prose

Here is the part most founders skip. When you get an answer from an AI model, do not focus on whether your company was mentioned. Focus on which sources the model cited to compose its answer.

In Perplexity this is easiest because the citations are right there. In ChatGPT and Claude with web search turned on, you can usually see the sources too. Look at what got cited and ask yourself a single question: do I have a presence on any of these sources?

What you will almost always find is that the cited sources fall into three buckets. There are publisher roundups and listicles ("Top 12 fractional CMOs for B2B SaaS in 2025"). There are independent reviews and comparison sites (G2, Capterra, Clutch, niche industry directories). And there are media mentions and expert commentary pieces (Forbes, industry trade publications, podcasts with transcripts).

If your competitors are showing up across all three and you are showing up in none, you do not have a content problem. You have a third-party endorsement problem. AI models trust third parties because third parties have less incentive to lie. Your own website saying you are the best fractional CMO in healthcare is worth almost nothing to a model. A roundup on a marketing publication's site saying it is worth substantially more, because the publisher had no commercial reason to flatter you.

This is the single most counterintuitive finding from running these audits, and the one that flips founders' instincts the most. Your own beautifully written homepage is not what is going to fix this. What fixes it is showing up in the places models are already reading.

Check what AI says about you, not just whether it knows you exist

The second cut of the audit is qualitative. After you have run the buyer-question queries, type your actual company name into each model and ask it to describe your business. Then ask it three follow-up questions. What are this company's main strengths. What are this company's weaknesses or limitations. Who are this company's competitors.

You will learn three useful things from this.

The first is whether the model has accurate basic facts about you, or whether it is hallucinating, or whether it has confused you with someone else. The Sarah situation I opened with falls into category three, and it is more common than you would think, especially for businesses with common names or names that overlap with a larger competitor.

The second is what narrative the model has constructed about your business based on the web sources it has read. This is your AI-mediated reputation. If a model describes your business in a way that surprises you, that is the version of your story most of your prospects are now hearing from their first research pass. That is worth knowing.

The third is who the model places you next to. If you run a premium B2B service and the model lists three discount competitors as your peer set, you have a positioning problem that is going to show up in your pipeline whether you address it or not, because the buyers using AI to shortlist vendors are getting that comparison baked into their initial frame.

Audit the off-site footprint, not just the on-site

The instinct after seeing weak AI visibility is almost always to rewrite the website. I understand the impulse and I have done it myself. The instinct is wrong, or at least premature.

On-site SEO fundamentals still matter and you should have them in order. Title tags, header structure, schema markup that describes who you are and what you do, clear service pages, accurate contact information, a real about page with credentials. These are table stakes. If they are broken, fix them, but understand that fixing them is not going to move the AI needle much on its own.

What moves the needle is the off-site footprint. Where else on the public internet does your business exist as a verified, indexable entity. I look at four things when I run this part of the audit. Third-party listicle and roundup inclusions in your category. Independent review-site profiles with actual reviews on them. Media mentions and expert citations, including podcast appearances where the show publishes a transcript. And business directories with verified information, weighted by which ones actually matter in your industry.

The asymmetry between on-site investment and off-site investment is where most businesses are upside down. I see founders who have spent forty thousand dollars on a website redesign and zero hours on getting included in the three roundup posts that come up every time a buyer in their category asks an AI model for recommendations. Reversing that asymmetry is usually the highest-leverage move available.

Diagnose the gap, then prioritize ruthlessly

Once you have run the buyer-question audit, read the citations, checked your own-name results, and looked at your off-site footprint, you will have a pile of findings. The temptation is to try to fix all of it. Do not.

Rank the findings by two questions. Which ones are showing up in queries with real buying intent. And which ones are the cheapest to fix.

A roundup post on a publisher's site that ranks for "best fractional CMO for healthcare" and currently does not include you is a high-priority fix because it touches a buyer-intent query and the cost to fix is one email and a follow-up. A general business directory profile that is missing or out of date is a low-priority fix even though it is easy, because the directory does not feed buyer-intent queries.

The version of this audit I run for clients ends with a ranked list of usually six to ten actions, with the top three labeled as the ones to do this month. The other actions stay on the list but they do not get touched until the top three are done. That discipline is what separates an audit that produces change from an audit that produces a document.

What to expect after you fix the top three

Two things, on a delay.

The first is that within thirty to sixty days of placement on a high-quality third-party source, you will start showing up in answers from at least one of the four models you tested. Usually Perplexity first because its indexing is fastest, then Gemini, then ChatGPT, then Claude. This is not a guarantee and the timing varies by source authority, but the pattern is consistent enough that I tell clients to expect it.

The second is that the model's narrative about your business will start to drift toward whatever language is being used about you on those third-party sources. If you got included in a listicle that frames you as the specialist in a particular vertical, that framing will show up in the model's description of you within a few months. This is why getting the framing right on third-party placements matters more than getting it right on your own website. Your own website tells the model what you claim to be. The third-party sources tell the model what you have been independently described as. The model trusts the second one more.

A note on what this audit will not tell you

This is a diagnostic, not a strategy. It will tell you where you are invisible and where the cheapest fixes are. It will not tell you whether your positioning is right, whether your offer is strong enough to convert the buyers who do find you, or whether the buyers finding you are the right ones. Those are different questions and they require different work.

But I would rather have a founder run this audit on their own than spend another quarter assuming AI visibility is something to figure out later. Later is already here. The businesses that get pulled into the new discovery layer in the next twelve months are going to have a structural advantage over the ones that get left out, and the cost to get pulled in right now is lower than it will be once everyone else figures this out.

If you run the audit and want a second pair of eyes on the findings, you can book a call here. I do these reviews with founders regularly and I am happy to tell you whether you are looking at a one-month problem or a six-month problem. Either way, you will know.

Linda Orr, PhD, is the founder of Orr Consulting, a fractional CMO and marketing strategy practice. Her work on AI-era marketing, the trust economy, and the Polish Paradox has been cited in Forbes and published in academic journals including the Journal of Business Research.