You can build your company’s visibility in AI answers without an agency. The open web holds separate tools and separate pieces of knowledge, but we have not found a methodology there that is assembled into one working order — so that order has to be built as you go, with every step tested on your own company.
The question, then, is not whether you can manage without an agency, but how much time the work will take you. To make the scale clear straight away: our baseline measurement is 10–15 questions a client actually asks, put to the four systems we work with — ChatGPT, Gemini, Perplexity and Google AI. That produces 40–60 answers to read, compare against each other and record, and it repeats every month, because a single measurement shows no movement.
An observation: SEO with one agency, AI visibility with another
We have an observation we consider important. Almost every owner who reaches us already pays for SEO and has it handled by another agency: the website is built carefully on the technical side, and the Google positions are there. Yet with the question about visibility in AI answers, those same owners come to us.
What they say is roughly the same every time: SEO seems to be working, the site is technically sound, but explain in plain words what AI visibility is and how it differs — and do that part. In other words, the owner senses the difference between the two layers of work before being able to name it. Where SEO ends and visibility in AI answers begins we separate in a different article; here it is enough to say that the two overlap without being the same thing.
What an owner can put in order without help
The first thing an owner can do unaided is put in order what is already written about the company in the sources the owner controls: the company website, the Google business profile, directories, and pages where the firm is mentioned. Nobody can do this part instead of the owner, because only the owner knows which services the firm provides today and which it does not provide and never did.
That order in your own data is not a theoretical matter is visible in the numbers of one of our clients. Out of 99 mentions of that company in AI answers, in 50 mentions the systems attributed to the firm services it had never provided. Separately from that, the AI kept repeating a service the firm had dropped years earlier — the systems were carrying forward outdated information found at some point in the sources about the firm.
A second case comes from the same client. On an authoritative platform there was a company page the owner had never created: the platform had generated it, put the client’s website address on it, and taken the business description from someone else. To an AI system that page looked official, so part of another company’s activity moved into the answers about this firm. An owner can find such a page and close it without help — but only when aware that it needs looking for.
That mentions of a company on third-party platforms carry weight at all is visible from the outside as well. In the Whitespark 2026 study, 47 local search specialists scored 187 factors, and that edition introduced a separate score for the first time: how much each factor affects visibility in AI answers. The study’s author, Darren Shaw, writes that “AI is revealing how important and foundational citations are”.
How to start the audit yourself
Before changing anything in the sources, look at what the systems answer about your company today. Start not with a set of systems but with your clients: ask the people who have already bought from you which AI tools they use, and build a small set of statistics out of their answers. Put your questions to the systems your clients actually use — our own set of ChatGPT, Gemini, Perplexity and Google AI took shape around local service businesses in Poland and may not fit another market.
Then ask the systems you chose the question your client arrives with: which companies would you recommend for this service in this city. Do not say your own company name in the question, or you will be testing not whether the system finds you on its own, but what it knows about you once prompted.
Ask those questions in a clean new chat, with no memory of earlier conversations, or the system will build its answer out of what you discussed with it before. ChatGPT has a temporary chat for this: the OpenAI help page states that in Unpersonalized mode such a chat “Does not use memory, custom instructions, or plugins”.
In the answers you get, look at three things: whether your company is among the ones named, which companies repeat from answer to answer, and what reasons the system gives for its choice. That is enough to tell whether the subject concerns you at all. How to run the full measurement we cover in a separate article.
Where doing the work alone hits a ceiling
The first ceiling is time. Those 40–60 answers have to be collected every month and then marked up, and the sources the systems cite have to be checked one by one. The same work takes us disproportionately less, because the method is well run in and the time it costs is predictable; for someone inside a company learning it from scratch, the same steps will take several times longer.
The second ceiling is reading the result. One and the same observation can be read in two ways: the systems do not know the company at all, or they know it wrongly. Those are different problems with different solutions. Add to that the fact that the systems move unevenly: in our data — three businesses, two industries, two cities, five measurements — the directions diverge, and the overall share of mentions can stand still while some systems name the company more often and others less. So the figures have to be read system by system rather than as one average.
Reading the result also covers the questions themselves. Choosing them in your client’s words rather than in the words of your own service offering is helped by having seen many cases across industries: we are in constant contact with the market and can see the words people actually use.
What is worth outsourcing
If your own measurement showed that the systems name other companies instead of yours, the first step can be outsourced. With us that step is an AI visibility audit: we check whether AI recommends your company and which competitors it names instead, and we show which sources the systems built that answer from. The scope and the price are on our services page. No one can promise that a given system will produce a given answer.
Does AI already recommend the company to clients?
We start with a diagnosis of the company’s current visibility and real customer decision scenarios.
VIEW THE AUDIT