You type "who builds decks in Prosper, Texas" into ChatGPT and get three company names. One of them is a competitor you know cuts corners. You are not on the list, and you have been doing this work for fifteen years.
If you are asking why AI recommends competitors and not you, the answer is almost never the quality of the work. An AI engine (ChatGPT, Perplexity, Google's AI Overviews, Microsoft Copilot) has never seen a job you finished. It names a business for one of four reasons: that business's website states what it does and where, in sentences the engine can quote; the web sources the engine trusts describe that business; its listings and reviews agree with each other; or it published the page that answers the exact question asked. Your competitor got at least one of those right. This post walks through each reason, how to tell which one is costing you, and what to fix first.
How an AI engine chooses the names it gives
When you ask an AI engine a "who should I call" question, it does not consult a ranked list of local businesses. It runs a web search behind the scenes, reads a handful of the pages that come back, and writes a short answer that names the businesses those pages describe. Three to five names, one line each, and no page two.
Two things follow. First, the engine can only name a business it read about in the last few seconds. Second, it describes the business the way the page describes it. If the page says "residential deck builder serving Collin County since 2010", that is the sentence that gets used. If the page says "crafting outdoor dreams", the engine has nothing to work with and moves on.
We measured this for our own trade in October. We asked three engines 120 questions about AI, website and automation help in Frisco, Celina and Prosper. Gemini cited business websites 340 times and directories four times. OpenAI's model behaved the same way. Perplexity cited business sites too, but also LinkedIn company pages 25 times and agency directories dozens of times more. Most businesses that were named appeared in one engine's answers only, because each engine has its own favourite sources. We were named in none of the 90 city questions, which is what started the work this post describes.
Reason 1: their site says it in a sentence, yours says it in a slogan
This is the most common reason. Open your competitor's home page and yours side by side and look for one plain sentence that says what the business does and where. "We repair and replace residential roofs in Frisco, McKinney and Allen." Most sites that get named have that sentence in the first screen. Most sites that don't get named have a hero image, a tagline and a button.
An AI engine reads text, not pictures. A city name that appears only in your footer map or your logo file does not exist to it. The fix costs nothing: write the sentence, put it under the headline, and repeat the pattern at the top of every service page.
Reason 2: the sources the engine trusts describe them, not you
Each engine has a set of sites it leans on for local questions. For most, that is business websites first. For some, directories and professional profiles matter almost as much. In our scan, Perplexity drew on LinkedIn company pages, DesignRush, Clutch and Sortlist in a way the other two engines barely did.
If your competitor has a complete LinkedIn company page, a profile on the directory for your trade and a listing with the chamber of commerce, and you have none of those, there are three more places where an engine can read about them and none where it can read about you. The directories differ by trade: a contractor's list is not a dentist's list. The quickest way to find yours is to run the questions yourself and write down which sites get cited. Our ten-minute test shows how.
Reason 3: their listings and reviews agree, yours contradict each other
Before an engine names a business, it is effectively checking whether the rest of the web agrees that this business exists, does this work, and is in this place. A Google Business Profile that says "Smith Plumbing LLC" with one phone number, a Yelp page that says "Smith's Plumbing & Drain" with another, and a website that lists a third is three weak signals instead of one strong one. Worse, the engine may blend you with a similarly named company somewhere else. One engine described our own company as offering services we do not offer, because it had merged us with a different business that shares part of our name.
Reviews work the same way. Engines read them and sometimes quote them. Thirty reviews that mention the service and the town ("fixed our water heater in Celina the same day") tell an engine what you do better than a page of adjectives. Five reviews that say "great guy" tell it almost nothing. The steadiness and the content matter more than the raw count.
Reason 4: they wrote the page that answers the question
Some questions have no good answer on your site because nobody wrote one. "How much does a tankless water heater cost to install?" "Can an AI assistant answer a dental office's phones after hours?" If your competitor has a 500-word page that answers that question directly and you have a services page that lists "water heaters" as a bullet point, the engine quotes the competitor. It is not choosing them over you. It is choosing an answer over a bullet.
Two companies in our scan were named for North Texas questions purely because they publish a page per city. Their offices are nowhere near here. The page was enough.
How to tell which reason applies to you
A fifteen-minute diagnosis. Do it in a fresh, logged-out browser window so the engine is not remembering earlier conversations about your business.
- Ask ChatGPT and Perplexity three "who does [service] in [town]" questions. Write down who is named and which sites are cited.
- Open your home page on your phone. Is there one sentence, before you scroll, that states your service and your area? If not, Reason 1.
- Look at the cited sites. Are any of them directories or profiles you do not have? If so, Reason 2.
- Search your business name on Google Maps, Apple Maps and Yelp. Do the name, phone and address match your website exactly? If not, Reason 3.
- Take the question the engine answered and look on your own site for a page that answers it. If the best you have is a bullet point, Reason 4.
Most businesses find two of the four. That is normal, and both are fixable without a redesign.
What to fix first
In this order, because the later items depend on the earlier ones.
- The sentence. Service and area, in text, on the home page and every service page. One afternoon.
- Consistency. The same name, phone, address and description on your site, your Google Business Profile, Apple, Bing, Yelp and your trade's directory. One afternoon, then ten minutes a month.
- Answer pages. One page per real customer question, with the direct answer in the first two sentences. One a week until the common questions are covered.
- Reviews that say what you did. Ask after every good job, and ask customers to mention the service. Ongoing.
- Re-check monthly. Same questions, same engines, count the mentions. The count is the only honest score; a position in one paragraph on one day is not.
AI engines don't judge the work. They quote what they can read, so the business that is easiest to read gets named.
How Deca Digital helps
We are a small studio in Celina, Texas, and this is the work we did on our own site after that scan. Our free AI visibility check runs the questions your customers ask through three engines and tells you which of the four reasons is costing you. If you want the fixes done, our services cover the site writing, the listings clean-up, the answer pages and the monthly re-check. If you want the plain-English background first, start with what answer engine optimization is.
Book a free discovery call on our contact page and we'll start with the check.