Find out what assistants say about you, word for word.
Your buyers are asking ChatGPT which supplier to use and acting on the answer. You cannot see those answers, you are not in the referral logs, and the first sign that you are losing is usually a quarter of missing enquiries.
The referral you never see is the one you cannot fix.
Search used to leave a trail. Somebody typed a query, clicked a result, and arrived on your site with the query attached. You could count it, and what you could count you could improve.
An assistant answers directly. It reads a handful of sources, names two or three businesses, and the person acts on that. If you were not one of them, nothing appears in your analytics at all. There is no line in any report labelled "recommended a competitor instead of you", and that is precisely the line that matters.
You cannot optimise for a surface you cannot see. The first job is making it visible.
Ask the questions your buyers ask, then read the answers properly.
- 1
Track the real questions
Drafted from your profile and from the queries Search Console already shows you, then editable. These are buyer questions in buyer language, not keywords.
- 2
Sweep on a schedule
Each question goes to each assistant, repeatedly, so the spread between runs becomes a reported range rather than a false precision.
- 3
Read who was named
You, your listed competitors, and the rivals you never listed but the assistants keep recommending. Position in the answer counts: first named is not the same as mentioned in passing.
- 4
Read what was said
Tone, whether the recommendation was unconditional or hedged, which sources were cited, and any claim about you that is simply wrong.
The questions you are losing, and what to do about each one.
Every line here describes something the software actually does. If one of them turns out not to be true of your setup, that is a bug and we want to hear about it.
Every lost answer becomes a brief you can act on.
A question where assistants named a rival and not you turns into a topic in your queue with one click, carrying the question, the rivals named and the sources they cited.
You see which sources the assistants actually trust.
The citation table lists the domains being read to answer questions in your category, whether each is yours, and how often. It is usually the fastest map of where to earn a mention.
It measures itself against a published standard.
The methodology panel grades the programme against the IAB's framework for measuring visibility in the AI era, taking the minimum across nine criteria rather than the average, and names the criteria holding the grade down.
It distinguishes a platform change from a market change.
When your share moves, the drift verdict looks at the shape of the change across the whole category. A model update that moved everybody is a different event from a competitor overtaking you.
Reputation is read from real pages or not at all.
The sentiment tab finds published reviews using a web-search-grounded engine, and discards any review whose host is not among the pages actually retrieved. It refuses to run rather than let a model without web access invent complaints.
What it does not cover, stated plainly
- Google AI Overviews is not measured. It is the largest answer surface in the market, and the product caps its own coverage grade because of it rather than quietly rounding up.
- Assistants answer differently to different people and in different places. What is measured is a controlled, repeatable question, not what any one buyer saw on any one day.
- An API answer and the consumer app are not identical. The app has memory, personalisation and its own system prompt. This measures the model's settled view of your category, which is the part you can influence.
- Entity disambiguation is not implemented. If a competitor shares your brand name, the sweep may credit them to you, and the methodology grade is capped for this reason too.
- With no sweep behind it, the screen shows a setup checklist. There is no sample mode and there will not be one.
Answers, Not Hedging.
See what they say about your business.
We will run the sweep against your domain before the call and bring the answers with us. Most of what people find surprising is on the citations table rather than in the score.