Home Free check How it works Guides Pricing About
[email protected] · Herne Bay, UK

Customers describe the problem, not your service. How does AI suggest you?

The short answer

When customers don't know your kind of service exists, they describe their situation to ChatGPT or Google AI Mode instead of searching for you. AI works out which kinds of fix suit that situation, then names firms it can describe confidently for it, often from pages it finds by searching the web. To be one of them, publish pages that describe the problem in your customers' own words, explain the options fairly, including yours, and get other sites to describe you in the same words.

Key takeaways

  • Customers of hard-to-explain services describe a symptom or situation. Your pages need to use those words, not only your service name.
  • AI decides what kind of solution fits before it names anyone. If your kind of fix isn't on its list, it can't suggest you.
  • Collect the real phrases from enquiry emails, first calls and reviews, then test them on the apps yourself.
  • Write one page per situation: the problem, who it happens to, the options, how you fix it, what it costs and where you work.
  • Ask the same problem questions on the same apps every month, and watch your enquiries.

1. Why customers describe the problem instead

Some services are a bit magic. You tell people what you do and they say "I didn't know you could do that". That reaction is a compliment, and it also tells you something awkward: before they met you, they had no idea your kind of service existed, so they could never have searched for it by name.

What they do know is their situation. Something has gone wrong, or has been wrong for a while, and they want to know what to do about it. So they describe it, often at length, with their own context: what happened, what they've already tried, what they're worried it will cost. More and more people put that description straight into ChatGPT, Google AI Mode, Gemini, Claude or Perplexity and ask for advice.

That's the moment your firm is either suggested or missed. If AI connects their situation to your kind of fix and to you, they arrive knowing what you do and why it fits their situation. If it doesn't, they try the obvious supplier first, or decide it can't be fixed, and you never hear about it.

2. What AI does with a problem description

For many questions about a real-world problem, the apps search the web before answering. OpenAI says ChatGPT search "typically rewrites your query into one or more targeted queries" that it sends to search providers. Google says AI Mode and AI Overviews "may use a 'query fan-out' technique", issuing multiple related searches across subtopics to build a response.

In practice, a problem description goes through roughly two steps:

  1. From problem to kinds of solution. AI works out what the situation is and which ways of dealing with it make sense: doing nothing, a do-it-yourself fix, the obvious big supplier, a specialist, and so on. Its searches at this stage are about the problem, so the pages it reads are pages that explain the problem.
  2. From kinds of solution to firms. For the options it settles on, it names firms it can describe confidently for this situation. It's more confident when it has read a clear description of what a firm fixes, and when other sites say the same thing.

So the firms that get suggested are the ones whose own pages and third-party mentions describe the problem in the words the customer used. A firm that is described only by its service name can be passed over even when it's exactly the right fit, because nothing AI read connected that name to this situation. Our guide on how to get recommended by ChatGPT covers the general mechanics in more detail.

3. Why "we offer X" gives AI nothing to match

Most specialist firms describe themselves the way they think about themselves: by the name of the service, the method, or the qualification. A homepage that says "we offer X, delivered by accredited specialists" is accurate. It just doesn't contain any of the words a customer uses before they know what X is.

When AI searches on the customer's behalf, it's searching in problem words. A page that never mentions the problem won't come up in those searches, and if it does come up through some other route, AI still can't tell that X is the answer to this particular situation. You may be the firm customers would choose if they knew you existed, and still be left out, because the link between the problem and your service only exists in your head and in your customers' memories.

This is also why bigger, more generic competitors often come up first. They've usually been written about more, and their pages explain the common version of the problem in plain terms. For the hard-to-explain version, there's often very little written down, which leaves room for a specialist who explains it well.

4. Find the words your customers actually use

You can't guess these. Specialists know the problem too well to describe it like a newcomer, so collect the phrases from the people who had it:

  • Enquiry emails and contact forms. Read the first two sentences of your last twenty or thirty. That's usually close to what the person would have typed into AI.
  • First calls. Write down how people open the conversation, before you've explained anything. Note the words they use for the symptom and what they think caused it.
  • Reviews and thank-you notes. Customers often describe the before: "we'd been living with it for years", "three other firms told us it couldn't be done".
  • The questions people ask before they know you exist. Ask recent customers what they were trying to fix, where they looked first and what they nearly did instead.
  • The apps themselves. Type those phrases into ChatGPT, Google AI Mode, Gemini, Claude and Perplexity, logged out or in a private window, and see which kinds of fix come up and which firms are named.

Group what you collect into a few distinct situations. Each situation is a candidate for its own page, and the exact phrases become your headings, your opening sentences and your test questions.

5. How to write a problem-first page

A problem-first page starts where the customer is, then works its way to you. It answers, in this order:

  • The situation, in the customer's words. Use their phrase as the heading and describe it in the first sentence, as they would.
  • Who it happens to, and why. The circumstances that cause it, so a reader (and AI) can tell whether it matches theirs.
  • What the options are, including ones that aren't you. Each with a sentence on when it suits and what it costs.
  • How you fix it. What you actually do, in plain steps, and how long it takes.
  • What it costs. A "from" price or a typical range. AI can't repeat "contact us for a quote".
  • Where you work. The towns, regions or countries you cover, stated plainly.

Here's a short worked example, using a made-up firm. Imagine a team of eight that fixes a problem most people assume can't be fixed. Their customers never search for the service. They say things like "it's been like this for years, is there anything that can actually be done?"

A service-first page for this firm would open with the service name and the team's accreditations. A problem-first page would open like this:

  • Heading: "It's been like this for years. Can it actually be fixed?"
  • First sentence: "Usually, yes. Most people are told it can't be fixed, or that the only option is to replace the whole thing, but in most cases a specialist can repair it in one or two visits."
  • Who it happens to: two or three sentences on the circumstances that cause it, and the signs that it's the fixable kind rather than the kind that does need replacing.
  • The options: live with it, try a do-it-yourself fix, replace it, or have a specialist repair it, each with when it makes sense.
  • How they fix it, the price and the area: the steps, a "from" price, the usual time it takes, and the counties they cover.

The second version contains the words a worried customer types, says plainly who fixes it, and gives AI facts it can repeat. Our guide on why your business isn't showing up on ChatGPT covers the other things to check on your site, such as whether AI can read it at all.

6. Explain the other options fairly

When someone describes a problem, AI rarely answers with one firm. It sets out a few ways of dealing with it and compares them. That comparison is where your kind of service either earns a place or gets left out, so it pays to be the page that explains the options well.

A simple table on your problem page does a lot of the work. For the made-up firm above, it might look like this:

OptionWhen it suitsWhat to watch for
Live with itIt's minor and not getting worseSome problems get more expensive to fix the longer they're left
Do it yourselfThe simple version of the problemCan make the fixable kind harder to repair
Replace itIt really is beyond repairUsually the most expensive and disruptive option
Specialist repairIt's the fixable kind and you'd rather keep what you haveFewer firms do it, so check where they work and what they charge

Be honest about when another option is the better choice. A page that says "if it's minor, you probably don't need us" is more believable, to people and to AI, than one that claims every case needs a specialist. Compare kinds of solution rather than criticising named competitors, and never call another firm bad. You can describe your own approach plainly and let the reader see the difference.

7. Get other sites to describe you the same way

Your own pages tell AI what you say about yourself. Other sites tell it whether to believe you, and they're also how it confirms which problem you're the answer to. If every outside mention calls you by the service name only, AI still has nothing linking you to the customer's situation.

  • Directory and profile descriptions. Rewrite your Google Business Profile, Bing Places, LinkedIn and trade body listings so the first line names the problem you fix, then the service.
  • Reviews. When you ask happy customers for a review, ask them to say what the problem was before they found you. That's in their words by definition.
  • Articles and quotes. When you're interviewed or write a guest piece, explain the problem and the options the same way your page does.
  • Partners who send you work. Ask them to describe you on their sites in problem terms ("if you've been told it can't be fixed, talk to...") rather than with a bare listing.

Consistency matters. When your site, your profiles and the articles that mention you all describe the same situation in similar words, AI can describe you with confidence.

8. Check it's working

Keep a list of ten or so questions written in your customers' words, mostly describing the problem, with a few asking who can help and a few comparing options. Ask them on the same apps every month, logged out, more than once each, because answers change from one go to the next.

For each answer, note three things. Does your kind of fix come up as an option? Are you named, and is the description right? Which pages does it cite? Change usually shows up in that order: first your kind of fix is described better, then your pages are cited, then you're named.

Then watch the number that matters. Ask every new enquiry how they found you, with "ChatGPT or another AI app" as an option, and notice whether people start opening with the problem phrases from your pages. If someone does tell you AI suggested you, our guide on what to do when ChatGPT recommends your business shows how to find out why. How it works explains how we run these checks for clients.

Sources

Questions owners ask

Most of our customers don't know our kind of service exists. Can AI still recommend us?
Yes, as long as it has read a clear explanation that links the problem your customers describe to the way you solve it. AI doesn't need the customer to know your service's name. It needs pages, on your site and elsewhere, that describe the situation in the customer's words and say who fixes it.
Should we stop using our service name on our website?
No. Keep it, because some customers do know it and AI uses it to file you correctly. Add the problem alongside it: lead each page with the situation in your customers' words, then name the service as the fix.
If we explain the other options, won't customers just go elsewhere?
Some will, and they're usually the ones who weren't a good fit. AI compares options whether you do or not, so it's better that it reads a fair comparison written by someone who knows the problem well. Saying plainly when another option suits better makes the rest of your page easier to believe.
Our customers all describe the problem differently. Which words do we use?
Use the phrases that come up most often in enquiries and first calls, and mention the common variations on the same page. AI matches meaning as well as exact words, so one clear page that covers the main ways people put it does more than a separate page for every phrasing.
How many problem pages do we need?
One for each distinct situation that brings customers to you, which for most specialist firms is a handful rather than dozens. Start with the situation behind your best enquiries and add more once that page is cited.
How long before AI starts suggesting us when someone describes the problem?
Your pages can start being cited within weeks of going live, but being suggested by name usually takes longer, often a few months, because AI wants to see the same description on other sites too. Ask the same questions every month so you can see it change.

Want to know what AI says when your customers describe the problem?

We'll ask ChatGPT, Google AI Mode, Gemini, Claude and Perplexity the questions your customers ask, starting from the problem rather than your service name, and show you which fixes they suggest and who they name, with dated screenshots and the first things we'd change. It's free and takes 48 hours. If you want more, a one-off audit and plan is £399 and monthly work starts from £799. See pricing.

Get your free check