Foldwren

How AI assistants find and recommend businesses

See in general terms where AI assistants get information about businesses, what you can control, and how to test what they say about yours.

Ask an AI assistant to suggest an electrician for a switchboard upgrade in your suburb and you will get a confident reply, often with business names in it. A fair question for any business owner: where did those names come from, and why not mine?

It depends on the tool, and the details are not public. But the general picture is clear enough to act on.

Where do AI assistants get their information?

Broadly, from two places.

What the model learned in training

An AI model is built by processing a very large amount of text, much of it from the public web. Along the way it picks up general knowledge, which can include well known businesses and widely discussed topics. This knowledge has a cut off date and is not updated day to day, so it can be stale. A small local business may barely feature in it at all.

What it looks up when you ask

Many assistants can also fetch current information while answering. Depending on the product, that might mean running a web search, reading pages from the results, or drawing on map listings, business directories and review platforms. When an assistant shows sources or links beside its answer, this is usually what happened.

For local recommendations, the second route matters most. That is good news, because it means current, public, well organised information about your business can be found and used.

Which exact sources each assistant favours, and how it weighs them, is something we cannot tell you. Nor can anyone else outside those companies.

Why does clear public information matter?

An assistant is assembling an answer from whatever it can read. Three things follow from that.

If it is not written down publicly, it does not exist. You may be the only electrician in town licensed for a certain kind of work. If your website never says so, no assistant will know.

Vague writing gives it nothing to use. "Quality service at competitive prices" cannot answer any question. "We upgrade switchboards in older homes across the Sunshine Coast" can.

Conflicting details create doubt. Two different phone numbers or sets of hours across the web make your information less reliable, and a cautious system may prefer a business whose details line up.

These tools also sometimes state things that are simply wrong. Clear, consistent information reduces the room for that, though it cannot remove it.

What can you control?

More than you might expect.

  • Your website. What it says, how clearly it says it, and whether the words are real text (text inside images and PDFs is harder for machines to use).
  • Your Google Business Profile and other map or directory listings: categories, hours, services, photos.
  • Consistency. The same name, address, phone number and description everywhere.
  • Freshness. Removing old offers, updating prices and hours, keeping the site alive.
  • Your side of reviews. You can ask customers for them and reply thoughtfully.
  • Whether your site allows access. Some sites block automated readers, sometimes by accident. Ask whoever looks after your site if you are unsure.

What can't you control?

  • Which sources a given assistant consults
  • How it words its answer, or whether it names anyone at all
  • What reviewers and other websites say about you
  • Whether the same question gets the same answer twice
  • Changes to the tools themselves

Accepting this list saves a lot of wasted effort. You are improving your odds, not operating a lever.

How do you test what assistants say about you?

Set aside ten minutes and do this.

  1. Open two or three AI assistants that you or your customers might use.
  2. Ask each one about your business by name: what it does, where it is, when it is open.
  3. Ask a recommendation question a customer might ask, such as who repairs hot water systems in your area.
  4. Write down anything wrong, missing or out of date.
  5. If the assistant shows sources, note which sites it drew on.
  6. Repeat the same questions a week later, since answers vary.

You are looking for patterns. A wrong phone number that appears repeatedly is coming from somewhere, and it is worth finding out where.

What should you do when an assistant gets it wrong?

Trace it back. Wrong details nearly always come from a public source: an old directory listing, a previous address on a social page, an outdated page on your own site that you forgot existed.

Fix the source and the correct information has a chance of flowing through over time. There is usually no button that corrects an assistant directly, and no guaranteed timeframe.

If you are not being mentioned at all, resist the urge to hunt for tricks. Look at the businesses that are mentioned. They tend to have detailed service pages, complete profiles, plenty of reviews and mentions on other reputable sites. That is the work, and it is the same work that helps you in ordinary search.

Your order