Key takeaways
- Start from the questions your buyers ask, not from your brand name. A competitor can only be recommended instead of you on a question you are asking.
- Ask each assistant each question more than once, and record who is named, who is recommended and which sources are cited. One answer is an anecdote.
- Read the rate across runs, for example named in 6 of 8 answers, and keep each assistant as its own series.
If you have seen an AI assistant name a competitor where you expected to see yourself, you have an anecdote, not a finding. Answers change from one run to the next and from one assistant to the next, so the only honest way to find out is to ask the same buyer questions many times and count.
This guide sets out that method in five steps, then lists the tools and providers that do parts of it, each described only from its own page.
Step 1: write the questions your buyers actually ask
Pick the questions a buyer would put to an assistant when they are working out what to buy. Cover the whole journey, from understanding the problem, to comparing options, to checking one vendor. Include questions that do not name any company, because those are where a shortlist is formed.
Freeze the list. A figure is only comparable over time if the questions behind it stay the same. Building a buyer-question set covers how to write and freeze one.
Step 2: ask every assistant, more than once
Run each question on each assistant you care about, and run it again. AI answers vary, so a single answer tells you almost nothing about how often a company is put forward. Two or three runs is the minimum. More runs give a steadier rate.
Keep each assistant as its own series. A rate for one assistant and a rate for another are never added or averaged.
Step 3: record who is named, who is recommended and what is cited
For every answer, write down three things.
- Who is named. Every company the answer mentions, yours included.
- Who is recommended. Whether a company is put forward as the choice, or as a named alternative, for the buyer's stated need. A name in a list of "popular options" is a mention, not a recommendation. Mention, citation and recommendation explains the difference.
- What is cited. The sources the answer points to, and whose they are. If a competitor's own pages, or a third-party page that describes them, keep appearing, that is the evidence the assistant is leaning on.
Treat an answer you could not read as missing. Do not count it as zero.
Step 4: read the rate across runs
Turn the record into counts. For one question on one assistant, you might see a competitor recommended in 6 of 8 answers and yourself in 1 of 8. Across a set of questions, the same count over a larger base is a steadier figure.
Two cautions. First, a rate needs its count and its base: 0.75 might be three of four answers or three hundred of four hundred. Second, look at where the gap sits. A competitor that is named in many answers but rarely recommended is in a different position from one that is recommended in most of them. Reading an AI visibility figure honestly sets out how to read a rate with its uncertainty.
Step 5: turn the gap into work
A question where a competitor is recommended and you are not is a piece of work. Look at what the answer cited. Often the competitor has a page that answers the question directly, or a source elsewhere that describes them, and you have neither. Evidence gaps turns a lost answer into a specific job, and the first ninety days puts the jobs in order.
Tools and providers that do this
You can run the method by hand on a small question set, and a spreadsheet is enough to start. Past a few dozen questions, most teams use a measurement tool for the repeated asking and counting. Tools measure; the work that changes the answers is done by people, either in your team or at an agency. Some teams pair the two.
The lines below say only what each company's own page says it offers and who it says it serves. They are not a ranking, and nothing here says one is better than another. Each carries the date its page was last checked.
Peec AIMeasurement tool
Its page says it offers AI search analytics: tracking brand performance on AI search platforms through visibility, position and sentiment, with competitor benchmarking. It describes itself as built for marketing teams.
Their pageLast checked
ProfoundMeasurement tool
Its page says it offers tracking of how often a brand appears in AI answers, analysis of what the answers say and which sites they cite, and benchmarking against competitors. It lists solutions for AEO, content, and PR and brand teams, and for agencies.
Their pageLast checked
ScrunchMeasurement tool
Its page says it offers tracking of how a brand shows up in AI answers, which sources are cited, and how often the brand appears compared with competitors. It lists use cases for enterprise brands and for agencies.
Their pageLast checked
RampiqAgency
Its page says it offers an AI search and GEO visibility audit for B2B brands, which includes a competitor visibility benchmark showing how AI treats competitors' mentions, with an action plan.
Their pageLast checked
Answer Position is one option too. It is a UK agency for B2B technology brands, serving UK and EMEA vendors, that measures where AI assistants name and recommend a brand on its buyer questions and does the marketing work that changes the answers. A free diagnostic looks at one question, and the Snapshot on our services page sets your brand against three competitors on a fixed question set.
What to do first
Write ten buyer questions, run each one three times on two assistants, and fill in who is named, who is recommended and what is cited. That is sixty answers and an afternoon of work. If it shows a competitor recommended where you expected to be, you have a first piece of evidence, and a list of the pages to look at next.
