When your buyers ask an AI assistant, does it name you?
More buyers now start with an assistant than with a search box. We measure what the assistant says when they ask, fix the reasons it doesn't name you, and measure again. The work is done by a system we built, checked by a person, and delivered as a report you can read in five minutes.
Measured, fixed, measured again.
Every finding carries the evidence it was found with. Nothing in the report is an opinion about your strategy.
Six questions your buyers ask
We run the questions a buyer would type — in your category, in your market — through an AI assistant with live web retrieval, and record who it named, whether your own site was cited, and where the answer was sourced from.
The reasons you weren't named
An assistant can only cite what it can read. We put the entity record, structured data, sitemap, headings and mobile correctness in place so there is something authoritative to cite — in your code, owned by you.
The same six questions, again
Thirty days on, we run the identical questions and show the difference. Not named, then named. That is the finish line, and you can check it yourself.
What being named is worth. What being absent costs.
The numbers below are other people's, published, and linked. The one number we will not give you is a dollar figure for your company. The report shows you which questions you are absent from and who is there instead; what that is worth is yours to judge.
A buyer who arrives from an assistant has already been told who you are.
Half of B2B software buyers now start their research in an AI chatbot more often than in Google: 51% in March 2026, up from 29% eleven months earlier. Of those, one in three bought from a vendor they had not heard of before the assistant named it.
The visits that follow are worth more. Across US retail sites, AI-referred visitors converted 42% better than the rest in March 2026, and Semrush values an AI-search visit at 4.4× an organic one. The buyer did the comparing before they clicked.
Being named is not traffic. It is being on the shortlist before the shortlist is written.
An assistant names a handful. When it does not name you, it names them.
In the reports we ran in September 2026, on every buyer question where a company was not returned, six or seven competitors were returned instead. The buyer was not left with no answer. They were handed someone else's.
69% of B2B software buyers chose a different vendor than they had planned because of what the chatbot said. If you are absent from the answer, you are the vendor they had planned, and this is how they leave.
It compounds quietly. 86% of what assistants cite comes from sources the brand itself controls, its own site first. A site with no entity record and no sitemap is not penalised. It is simply not there to cite, month after month, while the ones that are get cited again.
Nothing breaks. No alarm goes off. The buyer just never calls.
Sources: G2, The Answer Economy, 1,076 B2B software buyers, March 2026 · Adobe Digital Insights, Q1 2026 AI traffic report, US retail · Semrush, AI search visitor value, July 2025 · Yext, 6.8M AI citations, October 2025 · the six-or-seven figure is from our own September 2026 reports. The G2 figures describe software buyers and the Adobe figures describe US retail; we cite them because they are the largest published samples, not because they are your market.
What the report looks like.
A real Check from September 2026, reproduced as delivered, with the company, its domain and its suburb redacted. This is the first page: the result in one line, then the six searches exactly as typed and what came back.
Signal Collective · Share of Answer
Share of Answer: company
Of the searches a buyer would type before they know who to call, how many return the company.
The result in one line
Asked for commercial laundry suburb
, search returns company. Asked commercial laundry service Melbourne linen hire
, it does not.
We ran six searches on 8 September 2026 for commercial laundry in Melbourne: three the way a purchasing manager types them, one in company’s own words, one narrow and local, and one by name. domain was returned on three of the six. On the three buyer searches where it was not, between six and seven other commercial laundries were returned instead.
The six searches
What we typed, verbatim, and what came back
| No. | Search, as typed | domain | What came back |
|---|---|---|---|
| 1 | commercial laundry service Melbourne linen hire | Not returned | Six other commercial laundries were returned. domain was not among them. |
| 2 | towel laundry service gym Melbourne pickup delivery | Not returned | Six other commercial laundries were returned. domain was not among them. |
| 3 | restaurant linen hire Melbourne tablecloths napkins commercial | Not returned | Seven other commercial laundries were returned. domain was not among them. |
| 4 | the company’s own page title | Returned | domain was returned, at position 1. Seven other commercial laundries were returned alongside it. This search is built from the company’s own words, taken from its own page. |
| 5 | commercial laundry suburb | Returned | domain was returned, at position 3. Six other commercial laundries were returned alongside it. This is the narrow, local search: the one a company ought to win. |
| 6 | company name Melbourne | Returned | domain was returned, at position 4. Their Facebook page and a directory listing rank above their own site; own domain holds 5 of 10 results (positions 4, 5, 6, 8, 9); two unrelated similarly-named businesses also appear. This search is the company’s own name. |
The report continues: the pattern, the two causes, and what the site shows, with the evidence for each finding.
This Check was run through one search service, as its Basis line says. Every report states what was run, where, and on what date. The other businesses returned are never listed in the report.
Why most companies aren't named.
In September 2026 we measured nineteen company websites — four listed on the Singapore and Hong Kong exchanges, fifteen privately held in Australia. The gap was the same almost every time.
Counts are from our own scan of each site. Three of the four listed companies had no working mobile viewport; none of the fifteen private ones had that problem. Companies are not named here because none of them asked us to look.
Three things we sell. One of them is free.
Scope is agreed before we start and does not move. Whatever we put in your code, you own.
The Check
Six buyer questions run through an AI assistant with live web retrieval, plus a scan of your site for the reasons you weren't named. Delivered as a private report page.
- Who was named, in what position, sourced from where
- Every finding with the line of code it was found on
- No score, no grade, no invented dollar figure for what it's costing you
The Fix
We put the missing pieces in place — in your code, on your templates — then re-run the Check thirty days later and show the before and after.
- Entity record and Organization schema across templates
- Sitemap with dates, declared in robots;
llms.txt - Headings, meta descriptions, mobile viewport, https
- Anything with reputational or legal weight is checked by a person before it ships
The Watch
Twenty questions — the six from your check, plus fourteen more across your category — re-run every month, with a one-page delta: what changed, who moved, what to do about it. Cancel any time.
- The runs are automated; the reading is not — that is what the price is for
- Questions can be swapped as your market changes
- A person reads every report before you do
Prices are for one website in one language. Bilingual sites are scoped as two. We invoice in US dollars; Hong Kong dollars on request.
Start with the free check.
Five fields. We do the rest and send the report within five working days.
What happens after you press the button.
We write six questions from your phrase and your market, run them, and scan the site. A person reads the output before anything is sent.
The report page arrives by email. It is private, unlisted and yours to forward.
Reply to the email. We confirm scope in writing, do the work in your code, and re-measure thirty days later.
Keep the report. There is no follow-up sequence and nobody will phone you.
Signal Collective is a growth firm based in Hong Kong, working across New York and Sydney. Share of Answer is the productised part of what we do. If the report shows a bigger problem than a site can fix, we'll say so — that conversation is a scope call, not a form.