How to Get Your Business Recommended by AI
- Owen Measures

- 12 hours ago
- 6 min read
Someone in Medway needs a plumber. Five years ago they typed it into Google and picked from the map. Last year they still did. This year there is a decent chance they opened ChatGPT and asked "who is a good plumber near Rochester" and took the three names it gave them.
That is a genuinely new problem, because you cannot buy your way into that answer and there is no dashboard to log into. Here is what actually influences it, what does not, and what a small business can reasonably do about it.
First, the honest caveat
This field is about two years old and the people selling services in it, us included, know less than the confidence of their marketing suggests. The published figures vary wildly depending on who funded the study.
What follows is what holds up across multiple sources and matches what we see in practice. Where something is genuinely uncertain, we say so rather than dressing a guess up as a strategy.
The models do not work the same way
This is the part most guides skip, and it matters, because optimising for one does not automatically get you the others. Research across hundreds of millions of citations has found remarkably little overlap in which sites each model draws on.
ChatGPT leans towards established, reference-style sources. Wikipedia and similar authoritative sites make up a striking share of what it pulls from. It also tends to cite brands less often than the others, so getting named by it is harder but worth more.
Google's AI Overviews and AI Mode mostly promote pages that were already ranking well. If your regular SEO is sound you are in the running. If your site is on page four, you are not. This is the one where traditional SEO work translates most directly.
Perplexity heavily favours recent content and community discussion. Reddit turns up in a large share of its citations, and it weights freshness hard, so something published two years ago fades whether or not it is still correct.
Claude and Copilot sit somewhere between, both leaning on the same open web and both increasingly pulling live search results rather than relying on training data.
The practical upshot is that there is no single lever. What there is instead is a set of foundations that every one of them reads.
Your Google Business Profile is doing most of the work
For local service queries specifically, this is the single biggest thing, and it is free.
When someone asks any of these models for a recommendation near a place, the profile data is where the answer largely comes from. Your categories, your service list, your opening hours, your address, your description. If that profile is half filled in, you are not in the running, and no amount of clever website copy fixes it.
We wrote a full guide to getting your business on Google Maps which covers the setup properly. If you only do one thing off the back of this article, complete that profile.
Reviews are the strongest signal you actually control
AI recommendations are synthesised heavily from review data, and recent reviews carry more weight than old ones.
That is the uncomfortable part. Forty reviews from 2022 do less for you than eight from this quarter. Review gathering has to be a habit rather than a campaign you ran once when you set the profile up.
It also matters what the reviews say. A model reading "fast, tidy, turned up when he said he would" has something concrete to repeat back to whoever asked. A wall of five stars with no words gives it nothing to work with. When you ask for a review, ask people to mention what you actually did.
Reply to them too, including the difficult ones. That text is read as well.
Make your details identical everywhere
Your business name, address and phone number need to match exactly across your website, your Google profile, your Facebook page, directories, everywhere. Not roughly. Exactly.
Where those details disagree, models get less confident about recommending you, because they cannot tell which version is right. This is dull, unglamorous work and it is one of the highest return hours you will spend.
Check the obvious ones and then check the ones you forgot you had signed up to in 2019.
Structured data: telling the machine what you are
Schema markup is code that sits behind your pages and states plainly what your business is, where it operates, what it charges, what questions it answers. Humans never see it. Machines read it first.
Pages carrying proper LocalBusiness, Service and FAQ markup are meaningfully more likely to get pulled into AI answers, because the information is already in a form the model can lift without having to interpret your page layout.
Most small business websites have none of this. It is one of the clearest gaps we find when we audit a site, and it is a fixed piece of work rather than an ongoing cost.
Write the page that answers the question directly
Models quote text that reads like an answer. Not marketing copy, not a paragraph about your passion for excellence, an actual answer to an actual question.
That means pages built around questions real customers ask, with the answer in the first sentence rather than three paragraphs down. It means naming prices where you can, and giving specifics like areas covered, response times, what is included. Vague copy is unquotable, and unquotable copy does not get cited.
This is also why publishing genuinely useful content still matters. Not volume, not keyword stuffing, but a small number of pages that answer something properly and get kept up to date. Given how hard Perplexity in particular weights freshness, updating a good page beats writing a mediocre new one.
What does not work
Adding hidden text aimed at AI crawlers. Bulk publishing AI written articles that say nothing. Buying mentions on scraped directory sites. Anything sold to you as a guaranteed placement in ChatGPT, which is not a thing that exists.
Be sceptical of anyone offering to get you into AI answers for a fixed monthly fee with a guarantee attached. Nobody controls those outputs, and the honest version of this service is improving the foundations that influence them.
Where we stand on this
We include GEO work as part of our SEO service, which starts at £550 for the audit and setup and £250 a month if you want it kept up. In practice most of it is the same work that makes you findable on Google in the first place, done with an eye on how machines read your pages rather than just how Google ranks them.
The reason we bother mentioning it separately is that most small businesses are not doing any of it yet. That will not last. The businesses that sort their profile, their reviews and their structured data this year will be the ones being named when the rest catch up.
Frequently asked questions
Can I pay to appear in ChatGPT results?
No. There is no advertising product for AI recommendations at present. Anyone selling guaranteed placement is selling something they cannot deliver.
Does normal SEO still matter?
Very much, and more than the hype suggests. Google's AI answers mostly draw on pages that already rank, so good SEO is the foundation rather than a separate thing you replace. The cost of SEO has not changed much because the underlying work has not either.
How do I check whether AI mentions my business?
Ask it. Open ChatGPT, Gemini and Perplexity and ask each one for a recommendation the way a customer would, in your area, for your service. Do it a few times because answers vary. It is crude, but it tells you where you stand today.
How long does this take to work?
Profile and review improvements can show up within weeks. Structural work on your site takes longer, and it moves at the same three to six month pace as ordinary SEO. Anyone promising faster is guessing.
Is this worth it for a very local business?
Arguably more so. The number of businesses in any given town doing this properly is currently tiny, so the bar to being the one that gets named is far lower than it will be in two years.
If you want to know how your business currently looks to these tools, get in touch or call 01634 984929. We will run the searches with you and tell you honestly what we find, including if the answer is that you are already fine.



Comments