How AI apps decide who to recommend, and why it isn't you

Open ChatGPT and ask it the question your customers ask. "Who is a good plumber in Leeds?" Use your own trade and your own town. Most owners try this once, late in the evening, and end up looking at a list of three businesses that does not include them. One of the names is a firm they have never heard of. Another is smaller than they are.

It is a strange kind of bad news, because nothing has actually gone wrong. Your website is up, your reviews are good, and you might even be first on Google for that exact search. Nobody punished you and nobody outspent you. The app simply wrote a sentence about your trade in your town, and you were not in it.

The reason is duller than it feels

The app was not judging your work. It had to name two or three businesses out loud, and it carries the embarrassment if the answer turns out to be wrong, so it named the ones it could be sure about. You were not rejected. You were skipped, because there was not enough solid information about you to be worth the risk.

That sounds like a small distinction, and it changes everything about what you do next. If you had been judged and found wanting, the job would be to become a better business. It isn't. The job is to become an easier business to describe.

This stopped being a curiosity about a year ago

It would be reasonable to file all this under problems for later, and one number makes that difficult. BrightLocal asked 1,002 people in the US how they find local businesses, and the share who had used an AI app to do it went from 6% to 45% in a single year. That puts AI apps third as a source of recommendations, behind Google and Facebook, and ahead of the review sites most owners worry about far more.

It is one survey in one country, and a jump that large in twelve months deserves some caution before anyone builds a plan on it. Halve it, though, and the shape of the thing is unchanged. A channel that was marginal became ordinary while most of us were busy working.

Four reasons a business gets left out

The first is coverage. Your site says you work across "Leeds and surrounding areas", which sounds generous and tells nobody anything useful. Someone in a village twelve miles out cannot work out whether that includes them, and neither can anything else reading the page. The same problem turns up when the list of what you do lives inside a photograph, where no amount of reading will ever find it.

The second is hours. Your website footer says one thing, your Google profile says another, and a directory you have not logged into since 2019 says a third. None of them is obviously wrong, so none of them can be trusted, and a business that might be closed is a poor recommendation to make.

The third is that there was nothing specific to say about you. A page built around quality workmanship and customer satisfaction reads well enough to a human skimming it, but it contains no fact anybody could repeat to someone asking a concrete question. Pages like that are invisible in a way that has nothing to do with search.

The fourth is simply that a competitor was easier to describe. They wrote down their prices, their area and their hours in plain sentences, so they were the safe answer. Notice that three of those four reasons are about being clear rather than being good, which is the uncomfortable shape of this whole problem. The better business does not automatically win. The better described one does.

What to do this week, before you spend anything

Twenty minutes and no money. Ask the same question in ChatGPT, Gemini, Claude and Perplexity, and write down which businesses each of them names. Then ask each one directly about your business by name, and note down anything it gets wrong: old hours, a service you dropped years ago, a price that has not been true since 2023.

Finally, open your own website and try to find your service area, your opening hours and the jobs you do not take on. If you cannot find them quickly, nothing else can either, and most owners fix something real in that last step alone. It costs an afternoon and no more than that.

What will not help

An industry has grown up selling AI ranking services, and it is worth being careful with it. Google publishes a page about how its own AI answers work, and the wording is blunt:

There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.

Being found by search in the ordinary way is the whole requirement. So when somebody offers you a technique for getting into AI answers, the fair question is where that technique comes from, because the platform itself says there isn't one.

The real answer is much duller, and you already have everything you need for it. Write down what you do, where you work, when you are open and roughly what it costs, in sentences a stranger could repeat without hedging. That helps a customer reading your website at nine at night, and it happens to help every machine that reads it too.

This is the problem our own product exists for, so we think about it more than is healthy. hailera publishes an agent for a business that answers from what the owner has written down, so an AI app can ask it instead of guessing from a services page. You write the answers, you check them before anyone sees them, and it is free.

Sources

• BrightLocal, Local Consumer Review Survey — 1,002 US consumers; AI use for local recommendations rose from 6% to 45% year on year.
• Google Search Central, AI features and your website — on what is required to appear in AI Overviews and AI Mode.

Written by the team at hailera — we help small businesses answer customer questions without being there.