What AI visibility actually means, and how to measure yours in an afternoon

Ask ChatGPT "what is the best invoicing tool for a two-person agency" and it will name three or four products, say a sentence about each, and sometimes link to a page that backs the claim. If you make an invoicing tool and you are not in that answer, a buyer who trusts the model never learns you exist. That is the whole subject. Being in the answer is AI visibility.

It is not a ranking. There is no position two. You are either named, or you are not, and if you are named you are either the recommendation or the "you could also look at". The engines also cite: they link to a page that convinced them. Being cited is the stronger form, because the buyer can click through.

The four numbers that matter

Named rate. Of the questions your buyers ask, in what share of answers does your brand appear? Track it per engine, because ChatGPT, Gemini and Claude do not agree.

Cited rate. In what share of answers is one of your pages linked? This is usually much lower than the named rate, and it is the one that moves traffic.

Share of voice. Across all the names the engines mention for your questions, what share are you? This is the number that tells you whether you are a real option or a footnote.

Sentiment and accuracy. When you are named, what does the engine say? "Cheap but limited" and "the standard for small teams" are both mentions; only one sells. And sometimes the engine is wrong about you, which is the fastest fix there is.

Measure it by hand, this afternoon

You do not need software to get a first reading. You need an hour.

1. Write ten questions a buyer would ask before choosing you. Not your keywords: the questions. "Is X worth it for a small team", "X versus Y", "best tool for Z under a hundred dollars a month".
2. Ask each question in ChatGPT, Gemini and Claude. Fresh chats, no history.
3. Put the answers in a spreadsheet: question, engine, named yes or no, cited yes or no, the exact sentence about you, the other names.
4. Count. Named rate, cited rate, and the list of names that appear more than you.

Most founders who do this find two things. First, they are named far less than they expected, often in zero of ten questions. Second, the names that do appear are not the competitors they had in mind. They are the companies with the clearest pages.

What moves the numbers

Three things, in order of how fast they work.

Wrong claims. If an engine says you have no free plan and you do, fix the page it read. The answer changes at the next crawl. This is days, not months.

Pages that answer the question. The engines lift phrases from pages that answer the exact question asked. A page titled "Is X worth it for a two-person agency" with an honest answer gets quoted. A homepage with a hero line does not.

Mentions on the sites the engines trust. When an engine names three tools and links a review site, that review site chose the three. Get onto that page and you are in the answer.

Where a tool earns its keep

Doing the afternoon exercise once tells you where you stand. Doing it every day, across sixty questions and three engines, keeping every answer, and turning the losing questions into pages is a job. That is what ZwayRank's tracking does on a schedule, and what its crew does with the result. The first scan is free, so the afternoon can be replaced by a click, and you still get the spreadsheet.