AI doesn't recommend my business: 3 signs it doesn't trust you

When you catch yourself thinking "AI doesn't recommend my business", it's rarely an accident. It's a decision: the model had several options on the table and chose others. Not because the others are better businesses, but because they inspired more confidence at the moment it answered.
And there are nearly always symptoms first. The trouble is that nobody warns you.
When you lose positions on Google there are signs: traffic drops, impressions fall in Search Console, a light comes on somewhere. With AI there's none of that. There's no dashboard telling you "you were mentioned 4 times this week, your competitor 31". The conversation happens in a private window you can't see, and it ends with someone dialling a number that isn't yours.
It's probably the first time you can lose market share systematically without a single indicator showing it. There's no drop in traffic, because those people never reached your website. The phone just rings a little less, and you put it down to August.
These are the three signs we see again and again.
Sign 1: AI names you, but makes up your details
This is the cruellest one, because it looks like a win. You ask about your business and there you are. Your name. A reasonable description.
And an old phone number. Or the opening hours from before the refurbishment. Or the address of the premises you left in 2022.
Think about what that means. A customer who has already made up their mind, who wants to come to you, dials a number that isn't yours. They don't try again. They move on to the next one on the list. You never find out they existed.
When a model mixes your details with old ones or someone else's, it's telling you something: it doesn't have a version of you that's clear and repeated enough to be sure. It's filling in gaps. And filling in gaps, over time, turns into no longer naming you.
How to check (5 minutes). Ask about your business by name in ChatGPT and in Perplexity. Don't look at whether you appear — you will. Look at the phone number, the opening hours and the address, character by character.
Sign 2: you appear when they ask for your name, not when they ask about the problem
There are two very different kinds of question:
- "Tell me about Taller Riera, in Poblenou."
- "I need an e-bike repaired near Poblenou, where should I go?"
The first is a retrieval question: the model looks up what it knows about a specific name. Almost everyone passes it. The second is a recommendation question: the model has to choose. And that's where most fall down.
That's the gap between "existing" and "being recommended". Nobody opens ChatGPT to ask about a business whose name they already know; if they know it, they call. The questions that bring in new customers always start with a problem, a neighbourhood and a budget.
If you only appear when people search for your name, your visibility in AI isn't bringing you a single customer you didn't already have.
How to check (10 minutes). Write five questions the way someone who doesn't know you would write them, with the problem and the neighbourhood. Never your name. Note down which businesses come up.
Sign 3: the answer changes every time you ask
You try on a Tuesday and you're there. You repeat it on Thursday and you're not. Your business partner tries from her phone and three names come up that you'd never seen before.
It's tempting to read this as good news ("sometimes I show up"). It's the opposite. Instability is the fingerprint of doubt.
When a model has solid, consistent signals about a business, it names it in a boring, steady way. When the signals are scarce or contradict each other, the result turns random: whoever catches the wind best that day shows up. Being there "sometimes" means nothing is holding you in the answer, just chance.
How to check. The same five questions, on three different days, from a window where you're not logged in. If your name comes and goes, you have your diagnosis.
A warning about checking it yourself
Your account carries your history, your location and your previous conversations. The model already knows you, and it tends to tell you what you want to hear.
That's why the checks above are good for the initial scare and not much more. A single query from your phone isn't data. And it's exactly the screenshot the agency that calls you on Tuesday will show you: ask them how they measure, and compare it with the impossible promises of AI marketing.
On top of that, the answers aren't the same across assistants: out of 680 million citations, only 11% of the domains cited by ChatGPT were also cited by Perplexity. Showing up in one tells you nothing about the others, as we explain in why they don't recommend you the same way.
The context, in case you need a push
In Spain, use of generative AI went from 51% to 69% of the population in one year — eighteen points in twelve months — and frequent use, daily or several times a week, rose from 27.8% to 39.1%. The figures come from the second Observatorio Anual IAÓN (an annual Spanish AI survey), and they have the virtue of being Spanish, which is rare on this subject.
For finding local businesses specifically, the share using AI went from 6% to 45% in one year, according to BrightLocal — although that figure comes from 1,002 consumers in the United States, and it's worth saying so, because there's no Spanish equivalent yet.
There's no need to stretch it further. Neither figure says this is marginal, and neither says when you're going to show up.
Doing it yourself: the honest sum
An afternoon, free. The three checks above, plus claiming your Google profile and making your name, address and phone number identical everywhere. That fixes almost all of sign 1, and you don't need anyone.
A long weekend. Having a website of your own, readable, where those details live consistently. Without it, signs 2 and 3 don't budge: the model can't trust what it can't read.
And the part that never ends. Doing the three checks again. The answers change on their own without you touching anything, and your competitors move. Fifteen minutes every few weeks, forever — and it's the first thing people stop doing, because nothing reminds you.
What we do
That last part, above all.
Your wiweb page shows your Google profile live inside the page — hours, rating and reviews — so your details can't go stale in one place and not another: that's sign 1, closed at the root. And it's born with your services, prices and frequently asked questions written in blocks an AI can quote, which is what moves sign 2.
Sign 3 is the one that doesn't fix itself, and that's why we watch it for you. On Grow (€9/month) we check every month whether ChatGPT and Perplexity mention you, with the same questions each time, and which competitors appear where you don't. On Pro (€24/month) it's every week and across all four assistants, with up to eight competitors tracked. Always the same questions: if you change the question every time, you're not measuring, you're just looking.
Creating and publishing your page is free, no card needed: start here. Exactly what we check is in SEO and AI, and what each plan includes is on pricing.
What we don't do is promise you that an assistant will name you. Nobody can.
FAQ
If I do well on Google, shouldn't I do well in AI?
Not necessarily. Only 12% of what AI cites is in Google's top ten, and in ChatGPT it drops to 8%. We go into it in why your business is invisible to AI even if you show up on Google.
I have lots of good reviews. Isn't that enough?
They help, but they're not enough. We see businesses with hundreds of five-star reviews that don't appear in any answer, and others with far fewer that always come up. Reviews are one signal among several, and how fresh they are weighs as much as how many there are.
How long does this take to change?
Weeks or months, depending on where you start. Anyone promising you results in a week is selling you smoke and mirrors.
If my competitor appears and I don't, is it because they pay?
No. In organic answers there's no position for sale. Their advantage comes from accumulated signals, not budget — which is the bad news (you can't fix it with quick money) and the good news (neither can they).