The answer is built from other people’s pages.
An assistant asked for a recommendation fetches a handful of pages and writes its answer out of them. So the question that decides whether you get named is not what your own site says. It is which of those pages carry your name, and we can go and count them.
152 pages held the answer. Two named the business.
In August 2026 we took three businesses we work on, pulled the questions their customers ask, and recorded every third-party page ranking for those questions. That came to 152 pages: directories, roundups, community threads, supplier sites. We fetched each one signed out and searched it for the business’s name. Two of the 152 named them.
Two things follow. The pages an assistant would read while answering are mostly pages you are absent from, which is a gap you can close rather than a verdict on your business. And the absence is measurable, so the same count run again in three months says whether anything moved.
Three properties, questions that never mention the business by name, every page fetched signed out. Reported as a total on purpose: the per-business figures belong to those businesses.
A missing name and a wrong fact need different fixes.
When one assistant is quiet about you and another is not, the first thing we do is ask it what it thinks your address, hours and phone number are. If it answers with a former address or an old trading name, that is a data problem, and the fix is correcting the page it read.
If it has your details right and still names someone else, that is a sources problem, and the fix is the work above. The two look identical on a dashboard and cost very different amounts, which is why we run the check before commissioning a month of content against the wrong one.
Count, then close the gap.
Five things run every month, and the first two are what make the other three worth doing.
- The questions, fixed and repeated. A panel of the questions your customers ask, run across the assistants several times a month and averaged into one number.
- The pages behind the answers. We rank the third-party pages that come up for those questions, then read each one to see whether it names you.
- Wording you can submit. For the pages that take listings, we write the exact name, address, category and description for you to send, using claims you can back up.
- Wrong facts traced to their source. When an assistant states an old address or a former trading name, we find the page it read that from and get that page corrected.
- A page for a question you are losing. Each month we take a question where the answer names a competitor and publish the page written to get you named instead.
This runs inside the ongoing plan, $400/mo. You can also try one question free on the checker on the main page, which puts it to four assistants and shows you who each one named.
Questions about AI answers.
- Why does ChatGPT recommend my competitor and not me?
- Usually because the pages it read while answering name them and do not name you. An assistant does not hold an opinion about local businesses; it retrieves a handful of pages and summarises them. The work is finding which pages those are and getting you onto the ones that will take you.
- Can I pay to appear in an AI assistant's answer?
- No. There is no placement to buy in the answers themselves. What is for sale is the ordinary work underneath: being present and correct on the pages the assistants retrieve, which is why the job looks more like public relations and directory hygiene than like advertising.
- How do you measure whether an AI assistant names a business?
- A fixed panel of the questions your customers ask, put to the assistants repeatedly and averaged, so one lucky answer does not read as progress. We record two things separately: whether your name appears in the answer, and whether your site is used as a source, because they move at different times.
Find out which pages the answer is built from.
The audit ranks the pages that come up for your customers' questions, and tells you which of them name you today.
Perth, Western Australia