Does my brand appear in AI answers? How to check systematically
The usual check consists of typing in your own company name and then being pleased or annoyed by the result. That measures nothing. A usable procedure needs fixed questions, fixed conditions, and an evaluation that leads to tasks.
In short
- Asking about your company name measures nothing — you test with the questions your audience asks before they know you.
- Conditions have to stay constant: signed out, no history, same language, same wording, same rhythm.
- Four things get recorded: mention, position, competitors named, and the sources cited.
- The cited sources are the most valuable part of the evaluation — the concrete instruction follows from them.
There is no Search Console for AI answers. What there is, is a check you run yourself — and it is only worth anything if it is repeated under unchanged conditions.
Precision is not achievable here: two identical questions can produce different answers. Repeatability is achievable, and that is enough to see a direction.
Step 1: the question list
Twenty questions, unchanged for a year. The distribution decides the value.
| Type | Count | Example form | What it shows |
|---|---|---|---|
| Describing the problem | 8 | "Why do my emails go to spam?" | whether you count as an expert source |
| Method | 8 | "How do you set up SPF and DKIM?" | whether your guides land |
| Comparing providers | 3 | "Which providers are there for X in Switzerland?" | whether you are in the consideration set |
| The brand directly | 1 | "What does [company] do?" | whether the description is accurate |
Only one single question concerns your own name — and it is not there to measure visibility but to check accuracy. If you are described wrongly there, you have a different problem from a lack of visibility.
Step 2: clean conditions
- Signed out, or in a fresh window with no history and no stored personal details.
- Each question on its own, in a new conversation — not as a follow-up.
- Word for word as last quarter. Even small rephrasings change the result.
- Same language and region. If you work multilingually, check each language separately — the results differ considerably.
- Note the date. The systems change; without a date, a comparison is worthless.
Step 3: what to record
Four entries per question and system. More does not get filled in; less says too little.
- Mention: yes or no. Are you named in the text or in the sources?
- Position. Named in the running text, linked as a source, or both? A mention in the text weighs considerably more.
- Competitors named. Who appears instead of you — and is it repeatedly the same ones?
- Sources cited. Which pages serve as evidence: trade media, vendor pages, forums, comparison sites?
Worth knowing
The fourth column is the most valuable — and the one most often left out, because it takes the most work.
The cited sources show what kind of page the system considers citable. If two trade portals and one comparison service are cited consistently across your questions, that is an instruction: be present there, maintain an entry there, place a piece there. That often works faster than any improvement to your own site.
Step 4: evaluating
Three metrics come out of the four columns, and the tasks for next quarter come out of those.
Mention rate
The share of questions where you appear at all — separately per system. A low value is normal at the start; what matters is the direction across four quarters.
Text mentions
The share of mentions appearing in the running text rather than only in the source list. This number rises as your statements become more clearly extractable.
Competitor density
How many different providers appear across all questions? A narrow circle means: hard to get into, but valuable once you do.
Source pattern
The three to five pages cited most often. That list is the actual task list for the quarter.
A result that surprises many: often it is not your own page that is the problem, but your absence elsewhere. If a system consistently cites an industry directory, a trade portal and a comparison service for your questions, the fastest way into the answer runs through those three — not through your next piece of your own.
One maintained entry in a regularly cited source often works faster than three months of your own content work. That is unsatisfying, because it is not fully in your hands — but it is the cheaper order of operations.
Step 5: rhythm and effort
Quarterly, three systems, twenty questions: roughly two to three hours per pass. Checking more often gains nothing — the variation between two runs is larger than any change within a few weeks.
Continuously alongside that: reading the server logs. Which AI crawlers came, which pages they fetched, how often. That is the only figure in this area that is actually measured rather than observed.
I have run a visibility check across answer engines. Help me derive tasks from it. Starting point: - What we offer: [offering] - Audience and region: [details] - Systems checked: [list] - Period: [quarter] Results (per row: question | system | mentioned yes/no | text or source | competitors named | sources cited): [paste table] Tasks: 1. Calculate per system: mention rate, share of text mentions, number of distinct competitors. 2. Name the 5 sources cited most often, and for each tell me whether and how we could be present there. 3. Name the questions where we are closest – that is, already appearing as a source but not named in the text. For each, tell me what is most likely missing. 4. Name the competitors appearing consistently, and what they do differently in terms of content, as far as the data shows. 5. Derive five concrete tasks for next quarter, sorted by effect per effort. Do not invent figures or results that are not in the table.
In closing
Checking systematically means fixed questions, fixed conditions, four columns, quarterly. The result is imprecise and still usable, because it shows a direction — and because the column with the cited sources translates directly into concrete work.
The most common insight from the first pass is rarely the expected one. It is usually not "our content is too weak" but "we are absent from the places that get cited".
Common questions
How do you check whether your own brand appears in AI answers?
With a fixed list of twenty questions asked quarterly under unchanged conditions across several systems — signed out, no history, each question separately and worded exactly as last quarter. Record mention, position, competitors named and sources cited.
Why is asking about your own company name not enough?
Because that is not the question your audience asks. Anyone who knows your name will find you anyway. Visibility is measured with the questions someone asks before they know you — about the problem, about method, about comparing providers. The question about your own name only checks whether the description is accurate.
How often should you check?
Quarterly. More often gains nothing, because the natural variation between two runs is larger than any change within a few weeks. What can be watched continuously is the server log — which AI crawlers fetched which pages.
Why check while signed out?
Because signed-in accounts take conversation history and stored details into account. If you have mentioned your own company several times beforehand, you get an answer nobody else gets. The result then measures your own behaviour rather than actual visibility.
What do you do with the result?
The most important part is the cited sources: they show which pages the system considers citable. Presence in those sources — a maintained entry, a trade piece, a directory — often works faster than more content on your own site. The second most important part is the questions where you already appear as a source but not yet in the text.
Marketing that sets itself up
The Studio Engine beta is live. Claim your spot and help shape it from the start.
Join the beta →