AI audit
Two audits. Whether AI assistants are able to find, read and describe you at all, and whether the AI you have already put in front of customers is saying what you think it says.
What you get
Why this way
A business can be invisible to every AI assistant for two entirely separate reasons, and they are fixed in two different places by two different people. The first is robots.txt, a file on your own site naming the crawlers and telling them to stay out. The second is bot protection at your CDN or firewall, which returns 403 to those same crawlers no matter what robots.txt says.
We know the second one from our own estate rather than from an article. Opening one of our properties to AI crawlers took three changes: disabling a bot-protection rule, turning off a managed robots.txt that a platform had generated, and purging the cached copy of that file. Any one of those left alone kept the crawlers out while the other two read as correct.
Neither gate is visible from inside the business. Every page loads perfectly in your browser, the site ranks normally in Google, and nothing anywhere says that ChatGPT was turned away. The only way to find it is to ask the server as each crawler and compare, which is what the free audit does.
The second half of this work is the opposite problem. Plenty of shops now have an AI answering the phone or sitting on the website, and almost nobody has read a transcript of what it actually tells customers about price, timing or whether a repair is covered. That is a different audit and it is the one with real money attached to being wrong.
Run it yourself
The crawler half of this is built and live. The free audit at /audit/ checks both gates on any domain in under a second, reads only public records, and needs no account. Point it at your own site before you speak to anyone here. We ran it across all six of our own properties on the day it shipped, and we publish what it checks rather than describing it.
Run the free audit ↗Locally
The behaviour that makes this matter is local before it is anything else. Someone asks an assistant for a good body shop in Winter Park, or who does paint protection near them, and gets a short spoken answer naming two or three businesses rather than a page of ten blue links. There is no second page. A business that is not in that sentence is not in the running.
That is why this sits next to the local search work rather than apart from it. The same things decide both: a filed profile, a consistent name, address and phone everywhere, and pages that answer the question in the words a customer used. Being in Orange and Seminole counties means we can go and look at the counter, the phone and the listings in one afternoon, which is where the inconsistencies that confuse an assistant actually live.
Questions
No, and neither can anyone else. Nobody controls what a model says, the answers vary between users and change without notice, and there is no paid placement to buy. What can be done is removing the reasons you are excluded: being blocked at either gate, having nothing machine-readable that says who you are, and having no page that answers the question in the form it is asked. Anyone selling a guarantee here is selling something they do not have.
It overlaps and it is not the same. Search ranking is about being one of ten results a person chooses between. An answer engine returns two or three names in a sentence and the person usually takes one. The work is also different in shape: an assistant quotes a clear answer to a specific question, so a page structured to rank can be well-optimised and still never get quoted.
It is a real decision with a real trade-off, and it depends on what your site is. If your pages are the product, as with a publisher, blocking training crawlers is defensible and many publishers do it. If you are a shop that wants to be recommended, blocking them removes you from the answers. The usual problem is not the decision, it is that nobody made one: a platform default or a firewall setting did it silently.
Not directly, and the distinction matters. Google-Extended governs whether your content is used in Gemini and Vertex AI grounding, and it is separate from Google Search indexing, which follows Googlebot. So a site can block Google-Extended and rank normally in Search. The report names each crawler and what it actually governs rather than lumping them together.
What it claims about price, timing and whether something is covered, since those are the statements that create an expectation you then have to meet. Whether it tells the caller it is an AI. Whether recording is disclosed. What it does with a question outside its scope, because the failure mode is confident invention rather than silence. And whether a caller who asks for a person reaches one, which is the single most common gap.
Yes, and it is the more common case. We work from real transcripts and from testing it ourselves with the questions your customers actually ask, including the awkward ones. You get what it said, verbatim, and what we would change. If it is fine, the report says it is fine.
No. Certifying an AI management system against ISO/IEC 42001 is done by accredited certification bodies, and this house is not one. Anyone who offers you a certification without that accreditation is offering you a document that does not mean what you think it means. What we do is engineering and review work on systems that are actually running.
No, because there is not one. An assistant returns a sentence, not a numbered list, and asking the same question twice can return different names. Anyone reporting you as seventh in ChatGPT is quoting a figure that does not exist. What the report gives you instead is one sample with a date on it and the exact question printed next to the answer, so you can paste it in yourself and disagree with us.
Two questions. One about your business by name, to see whether an assistant knows you at all and whether the details it gives match what you publish. One without your name in it at all, phrased the way a customer would ask for your trade in your area, to see who gets named. The second is the one that matters, and leaving your name out of it is the only way the answer means anything.
Mostly in shape rather than in kind. The established tools are capable and they are enterprise software: an account, a subscription, a dashboard, and a monthly report aimed at a brand or an agency managing many clients. This is aimed at one owner who wants to know where they stand. The crawler check is free, takes under a second, needs no account, and the people who read it back to you are the ones who would fix it.
Taking new work
Send what you have and what is not working. You will get a straight answer about whether this house is the right fit, including when it is not.