A customer told you they found you through ChatGPT. Or worse: a customer told your competitor that, and you heard about it secondhand.

Either way, something has changed in how people find local businesses. They’re not always typing a search and scanning ten blue links anymore. They’re asking an assistant a question in plain language and getting back three or four names with a short explanation of why.

You can influence whether your name is one of those. Not perfectly, and not overnight, but far more than most owners think. This post walks through what we actually do to get cited by AI search, in the order we do it. Every step is something you can run this week and check for yourself.

How AI assistants actually decide who to mention

Here’s the part that changes how you’d spend your money: when ChatGPT or Perplexity answers a question about businesses in your town, it usually isn’t answering from memory. It runs a live web search first, reads a handful of pages, and writes an answer based on what it just read.

That’s a huge deal for you. It means you are not fighting a locked-in opinion the model formed a year ago. You’re competing to be in the small set of pages it pulls up at the moment someone asks.

Think of it like a new employee who’s been asked to recommend a supplier. They’ve got twenty minutes. They’ll Google it, skim the first several sources they land on, prefer the ones that state things clearly, and hand you a short list. They will not read your fifteen-page brochure.

So the job has two halves. Be findable when the assistant goes looking, and be easy to quote when it gets to your page.

Step 1: Find out what AI says about you right now

Before you change anything, you need a baseline. Otherwise you’ll spend three months on this and have no way to tell whether it worked.

Open ChatGPT, Perplexity, and Google’s AI results, and run these prompts. Use your real category and town, not “business” and “my city.”

  • “Best [what you do] in [your town]”
  • “Who should I hire for [the specific problem you solve] near [your town]?”
  • “Tell me about [your business name]”
  • “Is [your business name] any good? What do people say?”
  • “Compare [your business name] and [a competitor]”

Copy every answer into a document with today’s date. That document is your before picture.

Now read the answers for three things. First, do you appear at all? Second, if you do, is the description right — services, area you cover, price range, hours? Third, and most usefully, what sources did it link to?

That third one is the gold. Perplexity shows its sources next to the answer. ChatGPT links them when it searched. Those links tell you exactly which pages are shaping the answer about your industry. Sometimes it’s a directory you’ve never heard of. Sometimes it’s a Reddit thread. Occasionally it’s a competitor’s blog post.

One warning. Ask the same question twice and you’ll get slightly different answers. That’s normal. You’re looking for patterns across several runs, not one perfect result.

Step 2: Make your business details match everywhere

The most common reason an assistant describes a business wrong is boring: the facts don’t agree across the web.

Your website says you’re open till 6. Google says 5:30. An old directory listing has a phone number you dropped in 2021. A Facebook page you forgot about lists a suite number you moved out of.

A human customer shrugs and calls you. A model reading five sources at once sees a mess and gets cautious. It may hedge, describe you vaguely, or skip you for a business whose story is consistent.

Fixing this isn’t exciting, but it’s the highest-return hour on this list. Work through it in order:

  1. Pick the master version. Write down your exact business name, address, and phone number the way you want them everywhere. Decide on “St.” or “Street” and stick to it.
  2. Fix your own site first. Every other source eventually gets checked against it. Make sure the footer and contact page match your master version.
  3. Update your Google Business Profile. Hours, categories, services, service area. This is the single most quoted source about local businesses.
  4. Hunt down the stragglers. Search your phone number in quotes and see what comes back. Old listings, a defunct chamber page, a former partner’s site. Claim or correct what you can.

Then there’s schema. Schema is a small block of code on your site that spells your facts out in a format machines read directly. It labels your name, address, hours, services, prices, and reviews so there’s nothing left to guess at.

Your visitors never see it. It just removes ambiguity for anything reading the page. If you’re on WordPress, your SEO plugin probably generates some already; the question is whether anyone ever filled it in properly. It’s worth twenty minutes to check. This is standard practice in decent search engine optimization work, and if you’re already paying someone for SEO, ask them to show you what your schema currently says.

Step 3: Write pages in a shape AI can quote

Here’s the shift that took us the longest to accept. A page that ranks well and a page that gets quoted are not always the same page.

Ranking rewards depth and authority. Quoting rewards something narrower: a clear, self-contained answer the model can lift without reading around it.

So the structure that gets cited looks like this. Ask the real question as a heading, in the words a customer would use. Answer it immediately underneath, in two or three sentences, complete on its own. Then go deeper for the humans who want more.

That opening answer has to stand alone. If it says “as mentioned above” or “our process handles this,” it can’t be lifted, so it won’t be.

Three more things that make a page quotable:

  • Commit to specifics. “Most kitchen installs run three to five days” gets quoted. “Timelines vary based on your needs” gets skipped. Ranges are fine. Vagueness is fatal.
  • Put the facts in text, not in pictures. Pricing inside an image, hours inside a graphic, a service list built as a slideshow — none of it reads reliably. If it matters, write it as words on the page.
  • Date your pages and keep them current. A page that says it was updated last month gets more trust than one with no date at all.

The uncomfortable version of this: if you’ve been deliberately vague about pricing to get people to call, that strategy costs you AI mentions. You may still decide it’s worth it. Just decide it on purpose rather than by accident.

Step 4: The sources AI trusts that you’re probably ignoring

Remember the source links from step 1. For most businesses, they’re not your website.

They’re the places where other people talk about you. Assistants lean on third-party sources for the same reason a customer does — you’re the least neutral party about your own business.

The ones that keep showing up:

  • Your Google Business Profile and its reviews. Especially review text that names the specific service someone bought.
  • Community forums, Reddit especially. “Anyone know a good [trade] in [town]?” threads get read constantly. You can’t fake your way in, and shouldn’t try. But you can be genuinely useful in your local subreddit under your own name.
  • Local news and trade publications. One story in a paper your model already trusts outperforms a lot of blog posts.
  • Industry directories and association listings. The ones specific to your trade, not the generic link farms.

None of these is a trick. It’s the same reputation work that has always mattered, except now a machine reads it and repeats it to your customers.

One thing you can do this month: ask a few happy customers to mention the specific job in their review. “They rebuilt our online store and our checkout stopped dropping orders” is far more useful than “great service.” It gives an assistant something concrete to quote.

What doesn’t work: writing for robots

Every few years a new channel appears and someone sells a shortcut for it. This one is no different.

Stuffing your pages with phrases like “the best plumber in Waverly according to AI” doesn’t work. These systems aren’t matching strings the way search engines did in 2010. They’re summarizing meaning, and self-congratulation reads as marketing copy to be discounted.

Also skip the hidden-text tricks and the pages written entirely for machines. At best they get ignored. At worst they cost you with regular search, which still sends most of your traffic.

And here’s the honest counter-argument, because it deserves one: for a lot of local businesses, AI search is still a small slice of how customers arrive. If your Google Business Profile is a mess and your site takes eight seconds to load, fix those first. They pay off in both channels.

What makes this work worth doing anyway is that almost none of it is AI-specific. Consistent facts, clear answers, real reviews, honest specifics — that’s just a well-run web presence. AI search punishes the absence of it more visibly than Google ever did.

Check again in 90 days

Put a recurring reminder on your calendar for three months out. Then run the exact same prompts from step 1 and compare the answers to your saved baseline.

What you’re looking for, in order of how much it matters:

  1. Accuracy. Is the description of your business now correct? This is the first thing to improve and the easiest to verify.
  2. Presence. Do you appear in the “best [category] in [town]” answers where you didn’t before?
  3. Sources. Is your own site now among the cited links, rather than only directories?

Also start asking new customers how they found you, and give your team permission to write down the odd answers. “ChatGPT told me about you” is data, and right now it’s data almost nobody is collecting.

Expect this to be slow and uneven. Some answers will improve in weeks. Others won’t move until an outside source about you changes. That’s the nature of a channel built on other people’s pages.

If you’d rather see where you stand before doing any of this, we’ll run a free AI visibility audit. We run the same prompts from step 1 across the major assistants and send you a plain-English writeup of how AI describes your business today and what’s driving it. If your team also wants to get more out of these tools day to day, that’s what our AI tools training for businesses covers.

Either way, run the five prompts yourself this week. Whatever comes back is what your next customer is reading.