What AI answer-engine visibility is, and how to measure it

When someone asks ChatGPT "what is the best accounting software for a small agency?", they do not get ten blue links. They get a short answer that names a handful of products, explains why, and sometimes links to a few sources. If your brand is in that answer, you are on the shortlist. If it is not, the buyer may never hear of you.

AI answer-engine visibility is how often, and how prominently, your brand appears in those answers for the questions your buyers ask. The engines that matter today are ChatGPT, Gemini, Perplexity, and Google's two AI surfaces: AI Overviews at the top of search results, and AI Mode.

This guide explains how to measure it in a way you can trust and track over time.

Why you cannot just "check your ranking"

In classic search, a page has a position for a keyword, and that position is fairly stable from one hour to the next. AI answers do not work like that.

Answers vary between runs. Ask the same question twice and you may get a different list of brands, in a different order.
Answers vary by context. Location, language, whether the user is signed in, and earlier messages in the conversation all change the result.
Wording matters. "Best CRM for startups" and "which CRM should a five-person startup use?" can produce different shortlists.
There is no public rank to look up. Engines do not publish where you appear. You have to ask and read the answers.

So a single screenshot of a single chat proves very little. Visibility has to be measured the way you would measure anything noisy: by sampling many times and looking at shares and averages.

Step 1: Choose the questions your buyers actually ask

Your measurement is only as good as your question set. Aim for questions a real buyer types before they know which brand to pick:

Category questions: "best project management tool for agencies"
Problem questions: "how do I stop clients missing invoice deadlines?"
Comparison questions: "Asana vs Trello for a design team"
Segment questions: "best payroll software for a small nonprofit"

Leave your own brand name out of most questions. "Is Acme any good?" tells you how an engine describes you, but not whether it recommends you when the buyer has not heard of you yet.

Twenty-five questions is enough to start seeing patterns. A hundred gives you room to split results by product line, segment or stage of the funnel.

Step 2: Record more than "mentioned or not"

For every answer you collect, capture:

The full answer text, so you can go back and read it.
Every brand named, in order. The first brand named is in a stronger spot than the fifth.
The tone of each mention: positive, neutral, mixed or negative.
Every cited source: the URL and its domain.

That last item is easy to overlook and is often the most useful. Citations tell you where the engine got its information, which is where you can act.

Step 3: Turn answers into four numbers

Once you have a week of answers, four measures summarise them well.

Visibility share

Answers that mention your brand ÷ all answers analysed

If you track 25 questions on 2 engines once a week, that is 50 answers. If 12 of them mention you, your visibility share is 24%. This is the headline number: how often you make the shortlist at all.

Citation share

Answers that cite a page on your domain ÷ all answers analysed

An answer can mention you without linking to you, and it can link to you without naming you in the text. Citation share tells you whether the engine treats your own site as a source. Count any subdomain of your site, such as a blog or help centre.

Average position

The mean of your position across the answers that mention you

Position is your place in the list of brands the answer names. Only average over answers where you appear; otherwise an absence drags the number around in a meaningless way. A brand with a visibility share of 20% and an average position of 1.5 is in a very different situation from one with 60% and an average position of 4.

Competitor shares

Run the same calculations for three to five competitors. Visibility on its own is hard to judge; visibility next to the brands you lose deals to is easy to read.

Step 4: Sample sensibly

Because answers vary, how you sample matters as much as what you measure.

Keep the question set fixed from week to week. If you change the questions, you change the measurement.
Ask each question on each engine you care about. Engines draw on different sources and often disagree.
Repeat questions where you can. Asking each question twice per engine per week smooths out one-off answers.
Fix the location and language to match your market.
Measure on a schedule. Weekly is a good rhythm: frequent enough to see the effect of a change, slow enough that week-to-week noise does not dominate.
Exclude failures. If an answer did not come back or could not be read, leave it out of the totals instead of counting it as "not mentioned", and note when a week has noticeably fewer answers than usual.

Read the trend, not the week. A move from 24% to 27% in one week may be noise. A move from 24% to 35% over six weeks, after you published three comparison pages, is a signal.

Step 5: Look at the sources

Finally, rank every cited domain by how often it appears across the week's answers. This list usually contains a mix of review and directory sites, community threads, publishers and competitors' own pages.

For each frequently cited domain, ask:

Is there a page on this site about us at all?
If there is, is it accurate and up to date?
If a community thread is being cited, does it mention us?

The answers to those questions are your to-do list. That is the point of measuring visibility: not to admire a percentage, but to find the specific places where a buyer's question is being answered without you.

A simple starting checklist

Write 25 buyer questions without your brand name in them.
Pick the engines your buyers use, and a country and language.
Collect answers weekly, and keep the full text and citations.
Calculate visibility share, citation share and average position for you and your competitors.
List the most cited domains and check your presence on each.

You can do this by hand with a spreadsheet for a few weeks. If you want it done automatically, with recommendations and draft fixes, that is what CiteGaps is built for.