AI Visibility Audit
Your buyers ask ChatGPT, Perplexity, Gemini, and Google’s AI Overviews who they should be talking to, and those answers come back with specific company names on them. This audit measures whether one of those names is yours, benchmarks whoever is winning, and gives you a written plan for fixing it.
The problem
A client asked me a question recently that almost nobody was asking a year ago: when someone asks ChatGPT to recommend a company like ours, do we show up?
Buyer research has moved into AI assistants, and an assistant does not hand back ten links to compare. It hands back three or four company names with a sentence about each. By the time a prospect fills in your form, the field was already narrowed somewhere you cannot see. If you were not in that answer, you never entered the deal, and nothing in your analytics will ever tell you it happened.
What your buyer types
Who are the best companies for this, and who would you recommend I talk to first?
What comes back
Generative engine optimization, GEO for short, is the work of getting your company named in the answers AI assistants give. You will also see it called AEO or AI search optimization. Three labels, one job.
Search engines rank pages. AI engines make a judgment. The model reads what the wider web says about your company, weighs independent sources more heavily than anything on your own site, and then decides whether it can recommend you with any confidence. A company with a perfectly healthy Google presence can be missing from every AI answer in its category, which is what I keep finding when I run the test.
The other difference is where the fix lives. Most SEO work happens on your website. Most of what decides an AI answer happens off it.
What a bad result looks like
Their buyers are attorneys and insurance adjusters, people who pick an expert by asking someone they trust. So we tested whether the assistants were naming them at all.
Eight questions their buyers actually ask, run through four engines. Fresh session every time, and never the client’s name first, because a model will happily discuss any company you introduce it to. Measured June 2026, and AI answers are not fixed, which is why the audit records every run rather than reporting one score.
Thirty-two runs. They were named in none of them.
The useful part was in the engines’ own reasoning. On several runs I could watch the engine search the client by name, check their history and their reputation, weigh them against a competitor, and then leave them out of the answer anyway. They were not unknown. The engine looked directly at them and could not find enough usable evidence to put them forward. That is a different problem from nobody having heard of you, and it has a different fix.
What I do
I build the question set out of what your sales team hears on first calls, phrased the way buyers phrase it. Product names and brand searches are deliberately excluded, because those answers flatter you and tell you nothing.
Every question, in every major engine, in a fresh session with no leading and no hints. Each run gets recorded with what the engine said and why. You get the raw scoreboard, not a summary of it.
Somebody is getting named. I find out who it is in each answer, and more usefully, what the model says about them that it does not say about you. That sentence is where the whole diagnosis comes from.
Which lever is actually costing you the answer, and which work would change it. Ranked, so you know what to do first and what can wait.
What moves the answer
What independent sources say about you carries more weight than anything you publish yourself. Trade publications, industry associations, directories, review platforms, press coverage. The model reads outside sources as evidence and your homepage as a claim.
Plain sentences that state what you do, who you do it for, and what proof you have. A page titled “Conveyor Systems for Food and Beverage Manufacturing” can be quoted directly into an answer. A homepage promising to engineer tomorrow’s solutions today cannot be quoted into anything.
Whether a model can tell what you actually are. Consistent naming, one clear category, descriptions that agree across your site, your LinkedIn page, and every directory you appear in. Companies that describe themselves five different ways dilute their own identity.
Two disclosures, because overclaiming is how agencies get fired. Nobody outside these AI companies knows the exact weighting behind an answer, so treat the three levers as a well-supported operating model rather than a published algorithm. And this is not a schema trick. Structured data helps a machine parse a page, it does not give an engine a reason to trust you. Anyone selling an AI visibility fix that installs in an afternoon is selling you the afternoon.
The deliverable
The audit is a fixed piece of work with a defined end. You own the plan whether or not I ever touch it again. If you want it executed rather than handed over, I run that as an ongoing program and we can size it on the call. Nothing about the audit depends on you buying the second thing. Book an intro call and we will start with what your buyers are actually asking.
Fit
As far as I can tell, almost nobody in industrial and manufacturing is working on this yet. That comes from my own searching rather than a formal study, so hold it loosely. It does mean the category is unusually open right now, and the first company in a niche to become clearly citable tends to keep the answer for a while.
Straight answers
No. I cannot control what a model outputs and neither can anybody else. What I can do is measure where you stand today, explain why, and fix the inputs the engines are demonstrably reading. If an agency guarantees you a place in an AI answer, that is a good reason to stop talking to that agency.
They overlap, but the target is different. SEO is trying to rank a page in a list of links. This is trying to get your company named inside an answer that never shows a list. The heaviest lever, what independent sources say about you, sits outside anything an SEO plugin touches.
Yes, and you should before you pay anyone to do anything. Ask an AI assistant the questions your buyers actually ask, in a fresh session, and see whether your name comes back. It costs nothing and it tells you whether you have a problem worth spending money on.
Longer than you would like, and I am not going to quote you a number I cannot defend. The fixes on your own site are quick to make. Getting independent sources to say the right things about you takes months, and that is usually the lever doing the most work. This service is new enough that I do not have a clean median across enough clients to give you an honest timeline yet, and I would rather tell you that than invent one.
ChatGPT, Perplexity, Gemini, and Google’s AI Overviews as standard. If your buyers lean on something else, Copilot or an industry-specific tool, say so on the call and I will include it in the question set.
Then the audit says so, and the plan changes from getting into the answer to staying in it. Being named once is not a moat. The work becomes widening the gap between you and whoever is second, which is a better problem than the one most companies have.
Bring two questions your buyers actually ask. We will run them live on the call and you will see exactly what I see.
Book an intro call