There are two versions of this question and only one of them brings you work.
The first is “what does an assistant say when someone types my business name in.” That one is worth checking, and it has its own answer. But nobody types your name in unless they already know you.
The second is the one that matters: what happens when a stranger asks an assistant to find someone who does what you do. That is a shortlist, and shortlists are won differently than search results.
What is the person actually asking?
Almost never “who is the best marketing agency.” Real prompts carry a qualifier.
“Who does bookkeeping for construction companies in Ontario?” “I need an AI readiness assessment from someone who has worked in media.” “Find me a physio near Barrie who treats runners.” People type into an assistant the way they would brief a colleague, because it answers like one. The qualifier is the whole point of asking a machine instead of scrolling a directory.
That changes what you are competing for. You are not trying to be the biggest name in a category. You are trying to be the source that matches the entire request, qualifier included.
Why does it pick the businesses it picks?
Because one source answered the whole question in plain words and the others answered half of it.
Here is a live example from our own inbox. Someone in an HR role at an ad agency needed an outside consultant to assess AI readiness, and specifically wanted someone who understood the media industry. They asked ChatGPT. We came back, and a readiness call was booked within the week.
We are not the biggest AI consultancy in Ontario and we did not outrank anyone. What we had was both halves of that request written down in ordinary language, in the same place: this is the assessment we run, and this is the kind of business we run it for. The assistant was filtering, not ranking, and we were filterable.
That is the entire mechanic. Most businesses publish the first half and leave the second half in their heads.
What actually gets you onto the shortlist?
Four things, in rough order of payoff.
- Name the qualifier in plain text. Not “we serve a diverse range of clients.” Say the industries, the sizes, the towns, the specific situations. If you are good with dental practices and family law firms, those words need to exist on your site. An assistant cannot infer what you never wrote.
- Be one consistent entity everywhere. Same business name, same address, same phone, same description across your site, your Google Business Profile, directories, associations, and accreditation records. Assistants pattern-match across sources, and inconsistency is what lets a similarly named business elsewhere get mixed into your answer.
- Make sure the machines can read the page. A surprising amount of what businesses are proudest of is invisible to the crawlers behind these assistants, because it loads by JavaScript after the page arrives. Reviews and booking widgets are the usual casualties. That has its own answer too, and it is usually a quick fix.
- Be in the sources they already quote. For most categories the assistants lean on a handful of directories, association lists, and review platforms. You will not displace those in a broad search, so be listed in them instead. Low effort, and it puts you in the pool the answer is drawn from.
How is this different from SEO?
Ranking is a queue. You are trying to be higher up a list of ten blue links, and position one is worth more than position four.
A shortlist is a shorter list, and getting on it is about being quotable rather than being popular. An assistant assembling three recommendations wants a source it can lift a sentence from without hedging. One page that answers a narrow question cleanly can do that. Ten pages built around a keyword usually cannot, because none of them commits to a specific answer.
The practical consequence is cheerful for a small business: specific beats big here in a way it does not in traditional search. You are not fighting for the head term. You are being the only clear answer to a question the head term never covered.
How do I test whether it is working?
Ask the category question, not your name.
Open ChatGPT, Claude, Gemini and a plain Google search where an AI overview shows up. Use a fresh chat each time, so you see what a stranger sees rather than what your own history has taught it. Then ask three ways:
- The plain version: “Who does [what you do] in [your area]?”
- The qualified version, with the qualifier your best customers actually have: “Who does [what you do] for [type of business]?”
- The buying version: “I need [the outcome your customer wants]. Who should I talk to?”
Write down who comes back and, where the assistant shows its sources, open them. That list tells you two useful things at once: whether you appear, and which sources the machine trusts in your category. The second one is often the more valuable finding, because it is a to-do list.
Check again in a month. These answers move as sources move, and one reading only tells you where you stand today.
What if my category is crowded?
Narrow the qualifier until you are the obvious answer, then publish that.
Being the tenth plausible answer to a broad question is worth very little, because the assistant only names a few. Being the one clear answer to “who does this for businesses like mine, near me” is worth a great deal, because there is often no competition for it at all. Most businesses have never written that sentence down anywhere a machine can find it.
Start with the questions your customers ask you out loud, on the phone, before they buy. What it costs. How long it takes. Whether you handle their specific situation. Those are the prompts people type, almost word for word, and answering them properly on your own site is the one lever here you fully control.
Where to start this week
Run the three test prompts above and note who comes back. Then write down the qualifier your best five customers share, and check whether that phrase exists anywhere on your website. For most businesses it does not, and adding it is an afternoon.
If you would rather have the publishing side run properly than fit it around everything else, book a free readiness call. Twenty minutes, no pitch. We will run the prompts together and tell you which gaps are worth closing first.