How AI assistants choose which local business to recommend
Short answer: An AI assistant recommends the businesses it can find consistent, independent information about. It reads the open web, including business websites, directories, review sites and articles, and repeats what those sources agree on. No assistant publishes a ranking formula, so the practical work is making accurate facts about your business easy to find in many places.
People now ask an assistant “who is a good family dentist near me?” and get three names instead of ten links. Being one of the three matters. Much of what is written about how to get there is guesswork, so this article separates what is known from what is not.
What is known
Assistants answer in two ways. Sometimes from what the model learned in training, which can be a year or more old. Sometimes by running a live web search and reading the pages it finds. Most assistants now do the second for questions about local businesses, and many show the pages they used.
They read ordinary web pages. Your own website, business directories, review sites, news articles, lists such as “best plumbers in Denver”, and forums. You can see this in the sources an assistant cites.
They can be blocked. Each assistant’s crawler has a name, and a website can refuse it in its robots.txt file. A site that blocks them cannot be read by them.
What is not known
- No assistant publishes how it ranks businesses. Anyone who gives you a formula is guessing.
- Which data providers each assistant relies on. You will read claims that one assistant draws mainly on a particular directory or search engine. We have not found one backed by a named, repeatable study or confirmed by the company concerned.
- How stable the answers are. The same question asked twice can return different businesses. A single screenshot proves little.
What a local business can do
None of this is a trick. It is the work that makes a business easy for anyone, human or machine, to find and trust.
- State the facts plainly on your own site. What you do, where, for whom, your hours and how to reach you, in text, not only in images. One page per service and per location.
- Be the same everywhere. Identical name, address and phone number on Google, Bing, Apple Maps, Yelp and the directories for your trade. Conflicting details are a reason to leave you out.
- Have recent, real reviews. On more than one platform if your trade has more than one. Assistants quote what reviewers say.
- Be mentioned by others. A local news story, a trade association’s member page, a “best of” list. Independent pages count for more than your own.
- Answer real questions. A short article that answers something your customers ask (“how long does a crown last?”) gives an assistant something specific to quote.
- Let the crawlers in. Check that your robots.txt does not block the assistants’ crawlers.
How to measure it
Ask. Write down five questions a customer would put to an assistant, in the words they would use, with your town in them. Ask each assistant every week or two and record whether you are named, in what position, and which sources it cites. Look at the trend over a couple of months, not at one answer.
The sources matter most. If an assistant keeps citing a directory you are not on, or a list that leaves you out, that is your next task.
What to be wary of
- Anyone who guarantees a place in an assistant’s answers.
- Planted forum posts and fake “best of” pages. They get removed, and the account behind them with them.
- Text written only for machines. If a page would not help a customer, it is unlikely to help you for long.