How Many AI Answers Actually Name Your Brand? Counts, Not Just Rates
The Mention Ledger in Ayzeo's Citation Analytics counts exactly how many AI answers named your brand — "18 of 30 answers, 92 total mentions" — instead of a percentage that barely moves for a big brand and teaches nothing. This is a plain-language guide to reading it: why absolute counts beat average rates, and how the "What changed" drawer turns a flat trend line into next month's to-do list.
Table of Contents Expand
1. What "N of M AI Answers Name Your Brand" Actually Means
Ayzeo runs your saved prompts — the real questions your buyers ask AI assistants — against each AI model you have enabled: ChatGPT, Google AI Overviews, Claude, Perplexity, Gemini, Grok, and more. Every time a model answers a prompt, that is one AI answer. The Mention Ledger sits at the top of the Citation Analytics page and reduces all of that activity to one honest sentence. Say your dashboard reads (the exact figures will be your own):
18 of 30 AI answers name your brand · 92 total mentions
Read it literally, because it is literal:
- 30 is the number of AI answers Ayzeo collected in the selected period — your prompts crossed with the models you track (each prompt that ran on each model in the period contributes one answer).
- 18 is how many of those answers named your brand at least once. This is your answer-level coverage: in 18 of the 30 rooms where your buyers were asking, you were in the conversation.
- 92 is the count of every individual mention across all those answers. One answer can name your brand several times, so total mentions is always greater than or equal to the number of answers that mention you.
Two things underpin every number in this post, so it is worth stating once. First, the measurement basis: the ledger reads the latest run of each prompt on each model inside your selected time range — not an average of every run, but a snapshot of what AI would say today, bounded by the range. Second, the receipt principle: no metric without proof. Every figure is clickable and drops you onto the actual AI answer behind it, so "18 answers" is 18 real responses you can read, quote, and act on.
2. Why Absolute Counts Beat Average Rates
Almost every AI visibility tool leads with a rate: "56% mention rate", "visibility score 72". Rates are fine for a glance, but they fail exactly when the stakes are highest, for three reasons.
Rates hide the denominator. "50% mention rate" is one answer out of two on a tiny account and 400 answers out of 800 on an enterprise one. Those are completely different realities that collapse into the same number. A raw count never loses the denominator, because the denominator is printed right next to it.
Rates average away the movement that matters. A brand sitting at the #2 or #3 position in its market does not move its average rate much week to week, even when meaningful things happen underneath: it lost three answers on ChatGPT and gained three on Perplexity. The average looks flat; the business reality is not. Absolute counts, broken down per model, keep that movement visible.
Rates are not receipts. When a stakeholder asks "are we winning in AI or not?", a percentage invites an argument. A count with a drill-down ends it: click the number, read the answers.
A rate tells you the temperature. A count with receipts tells you which room is on fire and lets you walk in and look.
3. Breadth vs. Intensity: Why There Are Two Numbers
The ledger deliberately shows two numbers because they answer two different questions, and confusing them is the most common mistake we see.
| Number | Question it answers | What moves it |
|---|---|---|
| Breadth — answers that name you (18 of 30) | In how many buyer conversations do we appear at all? | Winning net-new prompts — being present where you were previously absent. |
| Intensity — total mentions (92) | How prominently do we feature where we do appear? | Being named repeatedly within an answer — depth of presence, not breadth. |
Breadth is fixed by getting into rooms you are not in yet (new content, new sources, new prompts you were invisible on). Intensity is fixed by becoming the answer's central recommendation rather than a footnote. A healthy account grows both, but if you can only move one this quarter, moving breadth is usually where the pipeline is.
4. The "What Changed" Drawer: Your Monthly-Report Generator
Next to each number the ledger shows a change chip — for example ↑ +2 on answers and ↑ +14 on mentions. This is not a vanity trend line. Ayzeo runs the exact same aggregation twice: once for your selected window, and once for the window of the same length immediately before it — disjoint, no overlap. Thirty days is compared to the thirty days before it; the UI tooltip shows you the exact dates on both sides.
Clicking "What changed" opens the drawer that writes your monthly report for you. It contains three things:
- A per-model breakdown — which AI moved. A gain on Perplexity and a loss on ChatGPT that net to zero on the headline are two separate stories, and the drawer keeps them separate.
- The biggest prompt-level changes — the individual questions whose mention count rose or fell the most, ranked by the size of the change and joined back to the prompt text.
- Mention flips — the sharpest signal of all, and the subject of the next section.
Instead of "mentions went up 14", you get "here are the four questions where you gained ground, the two where you slipped, and which model each happened on" — every line clickable down to the answer.
5. Mention Flips Are Your Actual To-Do List
A flip is a prompt whose outcome reversed between the two periods: a question where AI named your brand last period and does not now (a loss), or the reverse (a win). Because only prompts present in both windows can flip, a flip is never a measurement artefact — it is a real change in what AI tells your buyers.
The drawer shows each flip with a side-by-side view of both answers: last period's response next to this period's, on the same model, for the same question. When a prompt flips against you, you are usually looking at the exact sentence where AI now recommends a competitor in your place.
Don't report "still 78%." Report "these three questions flipped to a competitor last month — here are both answers." One is a number; the other is a decision.
This is the cleanest way to split the ledger into two jobs: the absolute counts are for reporting upward (a scoreboard a CFO can hold on to), and the flips are the to-do list (the specific prompts to go win back this month).
6. When Counts Can Lie: The Prompt-Set Warning
Absolute counts have one honest weakness, and the ledger is built to flag it rather than hide it. If you add or remove prompts between two periods, the count can move for a reason that has nothing to do with your visibility. Add 700 prompts and your counts jump — but that is portfolio growth, not visibility growth.
When the number of measured prompts differs between the current and previous window, Ayzeo shows an amber note under the ledger warning you that the delta may reflect a portfolio-size change, not a visibility change. It is a deliberate anti-vanity guardrail: we would rather tell you a number is not comparable than let it lie. This kind of methodological honesty is exactly what enterprise buyers respect — and, not coincidentally, exactly the "how to measure AI visibility correctly" content that answer engines like to cite. When you see the note, lean on the flips: they compare only the prompts present in both windows and are immune to the effect. The per-prompt breakdown, by contrast, still lists new and removed prompts (counting from or to zero), so read it with the warning in mind.
7. Turn the Ledger Into a Weekly Workflow
The metric is only worth as much as the habit around it. Here is the five-minute loop our most engaged customers run:
- Read the two numbers. Note breadth (answers that name you) and intensity (total mentions). Set the time range to match your reporting cadence — 7 days for a weekly standup, 30 for a monthly report.
- Check the deltas, then the warning. If the amber prompt-set note is showing, skip the headline delta and go straight to the flips.
- Open "What changed" and read the flips first. Wins → what worked, do more. Losses → open the side-by-side and read who took your place.
- Convert one flip into a task. Pick the loss that matters most to your pipeline and pair it with the Sources & Destinations drilldown to see which page won it. That is your content or digital-PR target for the week.
- Report the count, action the flips. In the deck, write "18 of 30, up from 15." In the backlog, write the three prompts you are winning back.
Do that for a quarter and the ledger stops being a chart and becomes the scoreboard your GEO program is actually managed against.
8. Frequently Asked Questions
- Q: What is the difference between "answers that name your brand" and "total mentions"?
- A: "Answers that name your brand" (e.g. 18 of 30) is breadth — how many AI answers included your brand at least once. "Total mentions" (e.g. 92) is intensity — every individual naming across all answers, since one answer can mention you several times. Total mentions is always greater than or equal to the number of answers that mention you.
- Q: What is a "mention flip"?
- A: A prompt whose outcome reversed between the two periods — AI named your brand last period and does not now (a loss), or the reverse (a win). The drawer shows both answers side by side, on the same model, so you can read exactly what changed. Only prompts present in both windows can flip, so a flip is always a real change, never a measurement artefact.
- Q: Why does Ayzeo show counts instead of just a mention rate?
- A: Rates hide the denominator, average away per-model movement, and cannot be clicked through to evidence. Absolute counts keep the denominator visible, expose real movement, and every number in the Mention Ledger drills down to the actual AI answers behind it. The rate is still available; the ledger simply leads with the count.
- Q: What time period do the change chips compare against?
- A: The window of the same length immediately before the selected range, with no overlap. If you select 30 days, the deltas compare it to the 30 days before that. The UI tooltip shows the exact dates on both sides.
- Q: Why is there sometimes an amber warning under the ledger?
- A: If a different number of prompts was measured in the two periods, a count can move for reasons unrelated to visibility — add 700 prompts and your counts jump because there are more questions, not more affinity. Ayzeo warns you when the measured prompt counts differ. Use the flips instead: they only compare prompts present in both windows, so they are immune to the effect.
- Q: Does the ledger count every run of a prompt?
- A: No. It uses the latest run of each prompt on each model within the selected window, and excludes failed or errored runs — a snapshot of what AI would say today, not an average of every run.
Key Takeaways
- Lead with the count, not the rate: "18 of 30 answers name your brand" keeps the denominator and the receipts a percentage throws away.
- Breadth vs. intensity: answers-that-name-you is breadth (get into more rooms); total mentions is intensity (feature more prominently in the rooms you are in).
- Every number is a receipt: it clicks down to the real AI answer, on the latest run per prompt and model in your range.
- Flips are the to-do list: prompts that reversed between periods, each with both answers side by side — report the counts, action the flips.
- The warning is a feature: when your prompt set changed, Ayzeo tells you the headline delta is not comparable instead of letting you report a phantom win.
See how many AI answers name your brand
Track ChatGPT, Google AI Overviews and more with counts, flips and receipts, not just a rate.