How does AI decide which brands to recommend?
There is no public formula, but when you analyse the answers three things keep coming up: what your website says, what others say about you, and whether it all matches.
First, the truth
Nobody outside the companies that build these systems knows the exact formula, and each assistant works differently. Anyone who promises they know it is exaggerating.
What you can do is observe. When you analyse many answers, three things keep coming up.
Where AI gets what it says
An AI assistant answers by combining two things: what it learned when it was trained, from enormous amounts of text, and what it looks up on the internet at the moment of answering. In both cases it is working with what has been published about you.
What your website says
Who you are, what you offer and for whom, explained clearly.
What others say
Media, reviews, guides and rankings that talk about you.
Consistency
All of the above telling the same story.
1. What your website says
This is the starting point. If your own website does not explain well what you do, it is hard for anyone else to do it for you. It helps when each service has its explanation, common questions are answered and the information is up to date.
The part you cannot see helps too: a tidy structure and labels that tell machines what each thing is, known as structured data.
2. What others say
This is the biggest difference from what was done before. According to McKinsey, in many cases a brand's own sites make up only 5 to 10 percent of the sources AI consults. The rest are third parties: media, specialist publications, user opinions and comparisons.
Source: McKinsey, “New front door to the internet: Winning in the age of AI search”, October 2025. Consumer survey.
Put another way: what you say about yourself weighs little against what others say about you.
3. Consistency
If your website says one thing, your Google profile another and an old directory a third, AI has doubts. And when in doubt, it tends to recommend whoever it can describe with confidence. Name, address, services, opening hours and description should match everywhere.
A simple example
Two clinics offer the same treatment. The first explains it in detail on its website, has a complete Google profile and appears in a guide for its sector. The second has a one-page website and different details in every directory.
Faced with the question “which clinic should I go to?”, AI has far more material to talk about the first. Not because it is better, but because it can describe it with confidence.
What you can do, and what you cannot
- You can make your information clear, complete and consistent.
- You can work so that reliable sources talk about you.
- You cannot decide what AI answers. Answers change from one day to the next and from one assistant to another.
That is why the first step is always to measure: ask the questions, see what happens and start from there.
Frequently asked questions
Can I find out why AI recommends my competitor and not me?
To a large extent, yes. By repeating the same questions you can see which brands appear and which sources AI cites. That lets you check where your competitors are and you are not.
Does AI get it wrong when describing companies?
Sometimes, yes. It usually happens when published information is scarce, old or contradictory. That is why it is worth checking what it says about you, not only whether you appear.
Do reviews have an influence?
User opinions are part of what AI can consult. We do not know how much weight they carry in each system, but they are one of the third-party sources.
Do all assistants give the same answer?
No. The same question can give different answers in ChatGPT, Gemini, Claude or Perplexity, and from one day to the next. That is why we measure across several at once.