How assistants decide who to recommend
When an assistant answers a local or a buying question it typically searches the web, reads the pages it finds, and composes an answer from the sources it judges reliable and relevant. Those sources are your own site, Google and Bing listings, review platforms, directories, and any articles or forum posts that mention you. If the assistant cannot read you, or reads three different versions of you, or finds nobody vouching for you, you are unlikely to be named.
That is the whole mechanism in plain English. The rest of this page is about making each of those sources say the same, clear, current thing.
The five causes
- Your site blocks or confuses the crawlers. A robots rule or a firewall that turns away AI search crawlers, pages built entirely in scripts the crawler cannot render, or a site that says nothing concrete on its main pages.
- Thin entity signals. Your name, address, phone, services and opening hours differ between your website, Google Business Profile, Bing Places, Apple Maps and the directories. To a model, that looks like several weak businesses rather than one strong one.
- Few recent reviews, or reviews that never mention what you do. Assistants lean on reviews to decide who is good at what. Twelve reviews from years ago that say "great service" do not tell it you fit implants.
- No answer-first pages. Your services page says "we offer a range of solutions". The competitor's says what they do, where, for whom, how quickly and what it typically costs, in the first hundred words. That is the page an assistant quotes.
- Nobody else mentions you. No trade body listing, no local news, no supplier page, no forum reply, no directory beyond the basics. Third-party mentions are how a model corroborates that you exist and are what you say you are.
Run the 20-prompt check yourself
Write twenty questions a customer would actually ask an assistant about your kind of business in your area: five about emergencies or urgency, five comparing options, five about cost or process, five naming a specific service. Ask them in ChatGPT and, if you can, in one other assistant. Record who is named each time and which sources the assistant cites.
You will end up with a list of the competitors who appear and the pages the assistants are reading. That list is your to-do list: it shows which sources matter for your trade and area, and what they say about the businesses that are being recommended.
What to fix, in what order
Most businesses have two or three of the five wrong, not all of them. The check tells you which, and the order above is the order that moves the needle fastest.
- First, be readable: check your robots file and firewall allow the AI search crawlers, and that your key pages have real text, not just images and scripts.
- Second, be consistent: one name, one address treatment, one phone number, one services list, across your site and every listing.
- Third, be current: a steady flow of recent reviews that mention the service and the place, and replies to them.
- Fourth, be quotable: rewrite the top service pages so the first hundred words answer the question a customer asks, with specifics.
- Fifth, be corroborated: trade-body listings, supplier and partner pages, local press, and a presence in the forums and directories that came up in your 20-prompt check.
We can run it for you
Send your business name and the town you serve through the form below. We run the 20-prompt check across the main assistants, record who is recommended and why, and tell you within one working day which of the five causes apply to you and what we would fix first. The check is free. If you want the fixing done, it is quoted as a fixed piece of work under the AI search optimisation service linked below.