It is not a ranking, and it is not random. Here is what is actually happening when an assistant answers who to hire, and what you can change.
It reads before it answers
An assistant asked for a local business does not recall you from training. It runs searches, opens some pages, and writes an answer from what it found in that moment.
So the question is not whether the assistant likes you. It is whether the pages it opened happened to mention you. That is a different problem, and it is one you can work on.
What it tends to read
- Roundup pages.“Best plumbers in” articles, local news lists, blog comparisons. These get quoted constantly.
- Directories and review sites. Whichever ones are big where you trade.
- Map and listing data. Names, ratings, review counts, categories.
- Your own site, if it can read it and the pages say something specific.
Notice that three of those four are not your website. This is why a good-looking site alone does not get you named, and why businesses with plain sites sometimes dominate these answers.
Being cited is not being named
There is a difference worth understanding. An assistant may list your website as a source at the bottom of an answer while never saying your business name in the answer itself.
A customer reads the answer. They do not open the footnotes. Being named in the sentence is the thing that gets you the call, and it is what we measure.
Where you sit in the answer matters
Being named third in a list of five is not the same as being named first. People take the first suggestion far more often than the last. So the useful question is not only whether you appear, but how early.
Why one check proves nothing
Ask the same question twice and you can get two different sets of businesses. We have measured this repeatedly. Asking once and seeing your name tells you almost nothing, and asking once and not seeing it tells you almost nothing either.
This is the single most misunderstood thing in this space. Any tool, ours included, that shows you a number from one answer is showing you a coin flip. The honest approach is to ask each question many times and report how often you were named, as a count rather than a tidy percentage.
It also means a zero is a ceiling, not a verdict. Not being named in ten answers means your real rate is low. It does not mean it is nothing.
What you can actually change
- Get onto the pages that already get quoted for your trade and your area.
- Make your own pages specific enough to be worth quoting from.
- Make your listings agree with each other, everywhere.
- Build the review count and rating that make you a safe recommendation.
- Let the AI crawlers read your site in the first place.
None of this is a trick. It is the same work that makes a business easy to choose for a person, done where a machine can see it.
Common questions
Can I pay to be named by ChatGPT?
Not for the recommendation itself. Ads are starting to appear beside these answers, and they are marked. The named businesses in the answer are chosen from what the assistant read.
Why did I get a different answer when I asked twice?
Because these systems are not fixed. The same question can return different businesses minutes apart. One answer is close to a coin flip, which is why a single check proves very little.
Does my website ranking on Google decide this?
It helps and it is not the whole story. Assistants often quote roundups, directories and review pages rather than the business site itself. You can rank well and still be missing from the pages an assistant reads.
How long does it take to change?
Fixes can go live in weeks. Whether answers move depends on how often the sources an assistant reads get refreshed, and nobody can promise a date for that.