Guide · AI search
How to get recommended by ChatGPT and AI search.
More customers are asking an AI assistant who to call instead of scrolling Google. The assistant reads the web, compares a few businesses, and names two or three. Here is what it reads, what we found when we tested it in Utah, and what to fix first.
The short answer
AI assistants recommend local businesses they can find on many trusted sources at once: Google and Yelp listings, review sites, the BBB, home-service directories, and the business's own site. To get named, be listed consistently everywhere they look, keep fresh reviews coming, and have a page that plainly answers "[service] in [town]." Schema markup and llms.txt help a little. Listings and reviews help a lot.
What we found when we asked AI assistants about Utah businesses
In September 2026 we asked four AI models with web search (from OpenAI, Google, Anthropic, and xAI, through their APIs) six questions the way a homeowner would. For example: "My water heater is leaking. Who's the best plumber near Cottonwood Heights, Utah? Give me 3 you'd actually call." Other questions covered HVAC in Sandy, an EV charger in Salt Lake City, roofing in Draper, garage doors in Murray, and landscaping in Holladay. Each question ran twice per model.
We got 35 complete answers that cited 869 sources between them. Three things stood out:
- Directories beat business websites. The most-cited sources were Yelp (68 citations), Birdeye review pages (42), the BBB (39), Angi (35), Thumbtack (25), HomeAdvisor (22), and Yahoo Local (20). The best-cited business website had 22.
- A few businesses win most of the answers. For several questions the same one or two companies were named across almost every model and run. Most businesses in each town never appeared.
- Any page that answers the question can get cited. Business pages titled like "Plumbers in Cottonwood Heights" were pulled in directly, and so were articles from companies that aren't directories at all.
Caveats. It was a small test: six questions, two runs each, through APIs rather than the consumer apps people use. Answers change from run to run. Treat it as a direction, not a law.
What AI assistants actually read
Every assistant with web search works roughly the same way: it turns the question into searches, reads the top results, and writes an answer from what it found. So the question is which results it sees.
| Assistant | Where local answers come from | What to do |
|---|---|---|
| Google AI Overviews, AI Mode, Gemini | Google's search index and Google Business Profile data. | A complete, active Business Profile and pages Google ranks. |
| ChatGPT | Its own crawler (OAI-SearchBot) plus third-party search providers. | Allow OAI-SearchBot, get indexed in Bing, be on the directories it cites. |
| Microsoft Copilot | Bing's index and Bing Places. | Bing Webmaster Tools and a claimed Bing Places listing. |
| Siri and Apple Maps | Apple Maps data, plus Yelp for reviews. | Claim your listing in Apple Business Connect. Keep Yelp current. |
| Perplexity, Claude | Their own crawlers and web search results. | Don't block PerplexityBot or Claude-SearchBot. Be on the cited directories. |
The checklist
In order of impact for a local service business:
- Claim and complete the big listings. Google Business Profile, Bing Places, Apple Business Connect, Yelp, BBB, Angi, Thumbtack, Nextdoor, and your local chamber of commerce. Use the exact same business name, phone, website, and service area on every one.
- Get reviews every week, on more than Google. Ask every customer after every job. Assistants read review counts, recency, and what reviewers say ("showed up same day," "fair price"). How to ask without breaking the rules.
- Publish a page for each service in each area you serve. "Furnace repair in Sandy" is a question. A page that answers it plainly, with price ranges and how to book, is what gets pulled into the answer. Only make pages you can make genuinely useful. Service-area pages that don't get penalized.
- Put the facts in plain text. Services, starting prices, hours, service area, license number, and how to book, in words on the page. Not in an image or a PDF.
- Add structured data. schema.org markup for your business, services, and FAQs, so machines don't have to guess.
- Let the crawlers in. Check your robots.txt isn't blocking OAI-SearchBot, Claude-SearchBot, PerplexityBot, or Bingbot. Many site builders and security plugins block them by default.
- Get mentioned locally. Local news, a supplier's "find an installer" page, a sponsorship, a neighborhood association. Mentions on sites the assistant already trusts carry weight.
- Add llms.txt. Ten minutes of work and harmless. Do it last.
What matters less than you'd think
- Keyword stuffing. Assistants summarize. Pages written for people get summarized well. Pages written for robots read like spam to both.
- Lots of thin pages. Fifty near-identical town pages can hurt you in Google, and assistants have no reason to cite them.
- "AI SEO" tools that promise rankings. Nobody controls what an assistant says. You can only control what it finds.
How to measure it
Write down the five questions your best customers would ask, with your town in them. Ask each one in ChatGPT, Gemini, Copilot, and Perplexity. Repeat a few times, because answers change. Count how often you're named and who's named instead. Then do it again next month.
That's the check we run for clients on the Found plan: the same questions every month, across the major assistants, with your share of mentions next to your competitors'. If you want a baseline now, ask for a free audit. AI visibility is part of it.
FAQ
AI search questions.
Can I pay to be recommended by ChatGPT?
Not in the organic answers. Some assistants are testing sponsored placements and partner integrations with directories, but the organic recommendation comes from what the assistant can find and trust about you on the web.
Does llms.txt get me into AI answers?
Probably not by itself. It's a cheap, harmless summary file, but no major AI company has said it uses llms.txt for ranking. Listings, reviews, and clear pages matter far more. See what llms.txt is and isn't.
Should I block AI crawlers?
Not if you want to be recommended. Blocking OpenAI's OAI-SearchBot, Anthropic's Claude-SearchBot, or PerplexityBot keeps your site out of those assistants' search results. A local business has nothing to protect by blocking them.
How do I check whether AI assistants recommend me?
Ask them the way a customer would, on several assistants, several times, and write down who gets named. Answers vary run to run, so one test proves little. Our monthly check runs the same questions every month and tracks your share against competitors.
Is this different from regular SEO?
Mostly it's the same work, weighted differently. Assistants lean harder on third-party sources (directories, reviews, local news) and on pages that answer a question directly. Good local SEO gets you most of the way.
See where you're leaking jobs.
Send your site. Within 48 hours you get a one-page teardown covering speed, search, Maps, reviews, calls, and AI visibility. No call required.