How do I get my local business found in AI search?

You run a plumbing company in Tacoma, or a dental practice, or a two-truck HVAC outfit. Somebody told you people ask ChatGPT for recommendations now instead of Googling, and that you need to do something about it. You do not have an SEO team. You have Monday morning and maybe three hours.

Good news and honest caveat in the same breath. Almost everything that helps is free, takes hours rather than months, and was worth doing anyway. But much of what you will read about local AI visibility is practitioner guesswork presented as method. This post marks what rests on evidence and what rests on inference.

Why local is a different problem than the rest of AI search

Local questions were always conversational. Nobody ever typed the way they actually think about finding a dentist — somewhere that takes Delta Dental and can see my kid on a Saturday — because search boxes punished that phrasing. Chat interfaces reward it, so local intent moved into chat early, ahead of the abstract software questions that get the attention.

The second difference is harsher. A results page gave you ten listings, a map pack and some ads, and plenty of businesses built a decent life in the middle of it. An AI answer names two or three businesses and stops. No page two, no map to expand. Being fourth means not existing, and that is new. The reason to care is not traffic — the traffic numbers are sobering, and we will get to them. It is that the shortlist got shorter.

Every engine gets its local facts from somewhere different

This matters more for local than anything else. On a general question, engines pull from broadly similar pages. On “plumber open now in Tacoma,” they reach for structured business data — and each reaches into a different pile.

Google behaves most predictably. AI Overviews and AI Mode sit on the same index and, for local questions, almost certainly the same business corpus that feeds Maps and the Local Pack. Google has not confirmed that for AI surfaces specifically, which is worth saying plainly. But the inference is strong, and Google's 2026 guidance agrees: “optimizing for generative AI search is optimizing for the search experience, and thus still SEO.”

ChatGPT leans on a Bing-derived search layer. Microsoft relaunched Bing Places for Business in October 2025 at bing.com/forbusiness — free, for appearing in Bing search and Bing Maps, with Copilot integrations flagged as coming. Whether ChatGPT reads those records directly is documented nowhere. What is clear is that being absent from Bing is a problem, and almost nobody in local services has claimed a listing.

Apple is two things at once now. Apple confirmed at WWDC 2026 that next-generation Siri runs on Google's Gemini, so the answering brain is Gemini's. But Apple Maps listings are still governed by Apple Business Connect, Apple's free tool for the place card shown across Maps, Siri, Spotlight, Messages and Wallet. Claim Business Connect for the map data; assume the conversational answer comes from Gemini.

Claude searches through Brave, whose index is independent of both Google and Bing — 30 billion-plus pages, roughly 100 million daily updates. If Brave has never indexed your site, you are not in Claude's answers, however well you rank elsewhere. Brave also says robots.txt is not what keeps a page out of its index, and it runs a submission form at search.brave.com/submit-url.

Step one: claim the three profiles

This is the first block of work and the least glamorous thing here. Three free listings, claimed and completely filled in — not claimed and abandoned.

1. Google Business Profile. Categories, exact hours including holidays, service area, phone, website, photos, attributes, full services list. Half-finished profiles are the local norm, which is why finishing yours is worth something.
2. Bing Places for Business, at bing.com/forbusiness. Free, and it imports your Google listing rather than making you retype it. Call it an hour.
3. Apple Business Connect. Free. Its place card appears across Apple Maps, Siri, Spotlight, Messages and Wallet. If your customers skew iPhone, that is no rounding error.

Step two: one name, one address, one phone number, everywhere

Then make your details identical everywhere they appear: the three profiles above, your website footer, Yelp, the chamber of commerce page, Facebook, the trade directory listing from 2014 you forgot about.

The mechanism gets explained badly almost everywhere, usually as a penalty. Nobody is docking you points; there is no consistency score. An engine assembling an answer has to decide whether “Harbor Plumbing, 1420 S Union Ave” and “Harbor Plumbing LLC, 1420 South Union Avenue, Suite B” are one business or two. When it cannot tell, it holds two half-confident records instead of one confident one. Confidence is what gets you named. So the fix is not perfectionism — it is picking one canonical version and using it verbatim. Street or St does not matter; using both does.

Be clear-eyed about the evidence. Consistency as an AI visibility lever is a practitioner claim, not a measured finding, and no engine has confirmed it. It is here because the disambiguation logic holds for any system that merges records, and it costs an afternoon.

Step three: put the real answers on your own site

Now the part most local sites genuinely lack. Your website has a Services page, an About page and a contact form. It does not have plain answers to what people ask before they pick up the phone.

• Which insurance plans you accept, by name.
• Whether there is parking, and whether it is free.
• Your holiday hours, written out, not “call for holiday hours.”
• Your service area as a list of towns, not a radius on a map image.
• What things cost — a range is fine, plus what moves it.
• Whether you handle the specific problem: sewer camera inspection, same-day crowns, wills for blended families.

Each of those is a passage an engine can quote. Generic reassurance is unquotable — there is nothing in it to lift. “We provide quality service to the greater Tacoma area” answers nobody's question. “We serve Tacoma, Lakewood, University Place and Fircrest, open until 6pm weekdays and 2pm Saturday” does.

Write them as the question where you can — a heading reading “Do you take Delta Dental?” with the answer underneath — and keep that answer in the first sentence or two, so it survives being pulled off the page.

Step four: reviews, and being talked about elsewhere

Ask for reviews from real customers, and reply to the ones you have. That gets a paragraph rather than a chapter, because most of what is written about review counts and AI is invented.

The one number worth citing comes with caveats. One aggregator analysis of AI citations found roughly 84% came from earned media — third-party coverage, directories, forum threads — rather than pages the brand owns, with only 5-10% from the brand's own site. That is a general finding across all query types, not a local one, and it comes from an aggregator rather than a study with a disclosed sample. Take the direction seriously and the number loosely.

The translation is unglamorous: the local news write-up, the neighborhood association page, the trade directory, the supplier's “find a certified installer” list. Those pages get crawled and quoted. Most of your citation surface is not your homepage.

Step five: structured data, honestly

Add LocalBusiness structured data with accurate hours, coordinates and links to your profiles. Twenty minutes, no cost, and still what Google reads for classic rich results and knowledge panels.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Plumber",
  "name": "Harbor Plumbing",
  "url": "https://harborplumbing.com/",
  "telephone": "+1-253-555-0142",
  "priceRange": "$$",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "1420 S Union Ave",
    "addressLocality": "Tacoma",
    "addressRegion": "WA",
    "postalCode": "98405",
    "addressCountry": "US"
  },
  "geo": {
    "@type": "GeoCoordinates",
    "latitude": 47.23874,
    "longitude": -122.47512
  },
  "openingHoursSpecification": [
    {
      "@type": "OpeningHoursSpecification",
      "dayOfWeek": ["Monday", "Tuesday", "Wednesday", "Thursday", "Friday"],
      "opens": "07:00",
      "closes": "18:00"
    },
    {
      "@type": "OpeningHoursSpecification",
      "dayOfWeek": "Saturday",
      "opens": "08:00",
      "closes": "14:00"
    }
  ],
  "areaServed": [
    { "@type": "City", "name": "Tacoma" },
    { "@type": "City", "name": "Lakewood" },
    { "@type": "City", "name": "University Place" }
  ],
  "sameAs": [
    "https://www.yelp.com/biz/harbor-plumbing-tacoma",
    "https://www.facebook.com/harborplumbing"
  ]
}
</script>

One rule above all: the hours in that markup must match your profiles and your footer, or you have added a fourth conflicting version of the truth.

What it is not is a proven AI-citation lever. Google states plainly that structured data is not required for generative AI features. As of early 2026 there were no peer-reviewed studies of schema's effect on LLM visibility, and neither OpenAI nor Anthropic nor Perplexity has said whether they parse it at all. Microsoft is the exception: Bing said in March 2025 that schema helps its models understand content for Copilot. Add it because it is cheap and correct, not because anyone promised citations.

How to tell whether any of it worked

You do not need an analytics team for this. You need a spreadsheet and half an hour a month.

1. Write ten questions a customer would ask before calling. Not keywords — questions with the local detail in them: “emergency plumber in Tacoma open on Sunday,” “dentist near me who takes Delta Dental.”
2. Once a month, ask those ten in two or three engines: Google's AI Mode, ChatGPT, and whichever third your customers use.
3. Log one thing per question — were you named, yes or no. Not your position.
4. Check analytics for referrals carrying utm_source=chatgpt.com. That is how ChatGPT tags outbound links, and your only first-party proof of being cited.
5. Read the trend across three months, not any one month.

What this is worth, and what to do Monday morning

Read the middle number carefully. Chatbots were under 1% of total publisher pageview referrals as of March 2026 — after ChatGPT referral traffic grew more than 200% during 2025. Growth that fast off a base that small is still a base that small. As a traffic channel, this is not one yet.

The honest reason to do this is better than the traffic argument. When somebody asks for a business like yours and gets two or three names back, you want to be one of them, and that shortlist is where local competition is heading. The cost of eligibility is an afternoon of free profile work and a few honest pages — nearly all of which pays for itself in ordinary local search whether the AI part materializes or not.

1. Fill in your Google Business Profile completely: categories, exact hours including holidays, service area, photos, services.
2. Claim Bing Places for Business at bing.com/forbusiness, importing your Google listing.
3. Claim Apple Business Connect and check the place card, especially the map pin and hours.
4. Pick one canonical name, address, phone and hours, then fix every listing that differs.
5. Write the six things your site lacks: insurance, parking, holiday hours, service area by town, price ranges, specific problems solved.
6. Add LocalBusiness JSON-LD with accurate hours, geo coordinates and sameAs links.
7. Ask your last twenty happy customers for a review, and reply to every review you have.
8. Put your ten questions in a spreadsheet, run them in two engines, log whether you were named, and repeat in thirty days.