Free tool · Brand perception in AI answers
AI is describing your brand right now. This is what it says.
Before a buyer reaches your site, they ask an assistant what you are, whether you are any good, and what else they should look at. You do not see those answers. We ask six of those questions, six times each, read every answer, and hand you the story the engine is telling, with the evidence and what would change it.
The problem
Your positioning is being rewritten by something that never read your deck
An assistant answering “what is this company” does not quote your homepage. It assembles a paragraph from review sites, forums, press and whatever it retained. That paragraph decides your category, your price bracket, your audience and your competitive set, and most brands have never read it once.
It is also consistent. Ask the same question six times and you mostly get the same story, which means the story is a fact about your market, not a fluke, and it can be measured and moved.
What is in the report
A narrative, then a plan
The story, written out
The account the engine settles on, in plain English, with the share of answers that tell it and the competing versions that did not win.
The qualities it assigns you
Every quality across every answer, counted, marked positive or negative, and flagged when your own site never claims it.
Who it offers instead
The products named beside yours, how often, and which of them the engine calls the stronger choice.
What it holds against you
Each objection quoted, counted and traced to the sources that carry it, with cautions true of your whole category kept separate.
Your site against the answers
Which of your published claims reach a reader and which never appear anywhere in the sample.
The moves
Ranked by what they cost against what they change, each with the counts that earned it and, where it applies, a snippet you can paste.
How it works
Measured, not guessed
You give us a domain, an engine and a country
We read the homepage and work out the brand name to ask about, then show you the exact questions before anything runs.
We ask six questions, six times each
Thirty-six answers from one engine in one country inside one window, so the counts describe one moment rather than a drift across weeks.
Every answer is read against a fixed schema
Positioning, stance, verdict, qualities, audiences, alternatives, objections and sources, recorded as verbatim quotes rather than impressions.
The numbers are computed in code
Not by a model. A model then writes the report over those numbers, and a validator rejects any figure it cannot find in them.
The six questions
- What is [brand]?
- What do people say about [brand]?
- What are the pros and cons of [brand]?
- Who is [brand] for?
- What are the best alternatives to [brand]?
- Should I use [brand]?
They are the questions a buyer types. We show them because a brand that writes better answers to these is simply doing its positioning work in public.
Who it is for
Anyone whose category is being explained by somebody else
Founders and marketing leads who want to know what the market is told about them. Agencies opening a conversation with a prospect: put your own name on the report and send the link.
You suspect buyers are comparing you with the wrong set of products.
Your pricing keeps being described from a page you do not control.
A competitor keeps coming up in calls and you cannot work out where from.
You have rewritten the homepage twice and nothing changed downstream.
Why you can trust the numbers
No figure in this report was written by a model
Every answer is coded against a fixed schema that permits booleans, verbatim quotes and names, and nothing else. Counts, shares and bases are computed in code from those codings. The model that writes the report is handed the numbers and may only put them into sentences; a validator then checks each figure, and each count-and-base pair, against the computed metrics and sends back anything it cannot find.
Where the report says the site does not claim something, that is from reading your pages in the same run, not from an assumption about what a site like yours usually says.
Questions
What people ask before they run it
Enough of the method to trust the numbers, and the questions themselves in full.
Is this the same as the AI Answer Audit?
No. The audit asks whether an engine knows you exist and gets your published facts right, from one question repeated. This asks what story it tells about you, across six different questions, and reports the narrative rather than the citations.
Why six questions and not one?
A single question gives you a definition. Perception is what a buyer assembles across the questions they actually ask in sequence: what is it, is it any good, what are the trade-offs, is it for me, what else is there, should I buy. The report is the composite.
Why ask the same question six times?
Answers vary between runs. One answer is an anecdote. Six lets the report say how consistent the story is, and consistency is the difference between a fluke and a fact about your market.
Does it cover every engine at once?
One engine per report, because answers differ so much between them that averaging would hide the finding. Run it again on another engine to compare.
Do you need access to my site or analytics?
No. We read your public pages the way a crawler would, and we ask the engine the way a buyer would. Nothing is connected and no login is needed.
Will a report be public?
Reports are not indexed and are not listed anywhere. Anyone with the link can open it, which is what makes it easy to send to a colleague or a client.
Read the paragraph your buyers already read
Free, no email needed, about four minutes.
Run it on your brand