Sample report
Share of answer for one consultancy: named in 88 of 133 answers, across nine passes on three assistants.
Every figure below is from one study-abroad cycle: 133 answers, three repeats of each question on each assistant, run on 17 September 2026. Nothing is illustrative. The companies are withheld because we do not publish a firm's report without asking them first.
gpt-5.6-lunagpt-5.6-terragemini-3.6-flashCycle 2026-09-17 · 20 questions scored · 3 passes each
Section one
How often the firm was named when a customer asked about the category. We show every pass, because a single average would hide that one assistant never named them at all.
Two passes named the firm in nothing at all. Any report that gave you one number for this quarter would be quoting a figure that never actually happened.
Section two
The same question, the same week, asked on three tiers people actually use. The tier mattered far more than the run did.
Range across passes: 55% to 83%
Range across passes: 0% to 40% — named in one pass out of three
Range across passes: 70% to 80%
The gap between assistants is 80 points. The gap between repeats of one assistant is 40. Which assistant your customer opened decided the answer roughly twice as much as luck did — so the fix is specific to a platform, not a general marketing push.
Section three
Being mentioned is not the same as being recommended. These are the 88 answers that did name the firm, counted by the position it held in the list.
Average position 3.4, and first place exactly once in 133 answers. The firm is a name the assistants add to a list, not the one they lead with.
Section four
Nobody supplied a competitor list. These are the names the assistants produced on their own, which is usually a different set from the one a firm thinks it competes with.
| Named in answers | Share | Avg position | Visibility |
|---|---|---|---|
| Category leader | 61% | 1.3 | 54 |
| This firm | 66% | 3.4 | 22 |
| Rival — platform | 28% | 2.9 | 12 |
| Rival — national chain | 32% | 3.7 | 10 |
| Rival — regional specialist | 14% | 1.9 | 9 |
The firm is named more often than the leader and is worth less than half as much. Visibility weights position, so being third on almost every list scores below being first on fewer. The regional specialist is named in a seventh of the answers and nearly matches the national chain, because when it appears it appears near the top.
Section five
Every page the assistants cited was fetched and read for who it mentions. This is the part of the report that turns into work.
Those 32 pages are listed individually in the report, ranked by how many answers each one shaped and split into two piles: a rival's own site, which you will not get onto, and independent pages — directories, forums, university and press pages — where a correction or an inclusion request is a realistic week of work.
The 13 flickering questions are the cheapest ground to take. The firm already appears in some passes of each, so the position exists and is not holding. Consolidating those is a smaller job than winning a question where it has never been named once.
Section six
The noise floor in this cycle was 58%. Ask the same questions twice, minutes apart, and well over half the named brands change. Every movement smaller than that is reported as flat, including the flattering ones.
Answers produced without a live search are separated out and never scored. They describe what a model memorised during training, not what your customer is told today.
Question weights are our own judgement, stated in the report and open to argument. Nobody publishes search volumes for AI questions, so anyone showing you one has invented it.
Nothing here connects visibility to revenue. No honest method does that yet.
Your turn
Send your company and category. We run the first report and send it whether or not you become a customer — and we will not publish it.