AI search for home services: how homeowners ask ChatGPT for a contractor, and how to be the one it names.
A homeowner in Liverpool, New York, woke up to a stain on the ceiling last week and did something that would have been strange three years ago: she described the problem to ChatGPT and asked who to call. It gave her three roofers, a line on each, and the sources it used. Another homeowner, planning a furnace replacement for next spring, asked Google and got an AI Overview with a price range and four company names before the first regular result. A third asked Perplexity which bathroom remodelers in the county had the best reviews.
Different engines, same pattern: the homeowner gets a shortlist before they ever see a website, and the shortlist is built from what the engine can find and confirm about each company. This guide is about getting on it. It's written for roofers, HVAC and plumbing companies, electricians and remodelers, and it assumes you'd rather fix things than read about them.
What homeowners actually type
The prompts fall into four kinds, and they want different things from you.
| Kind of prompt | Example | What the engine looks for before naming you |
|---|---|---|
| Emergency | "Furnace died, it's 10°F, who can come tonight near Cicero?" | Hours (including after-hours), service area, a phone number, recent reviews mentioning speed |
| Planned job | "Best roofers in Syracuse for a full replacement, and what should it cost?" | Price ranges on your site, project pages with photos, manufacturer certifications, review volume and recency, licence and insurance |
| Vetting | "Is [your company] legit? Any complaints?" | Consistent facts across your site, Business Profile and listings; BBB and review platforms; the licence database; how you respond to bad reviews |
| Comparison | "[You] vs [competitor] for a heat pump install" | Which of you states specifics: brands installed, warranties, financing, years in business, what's included |
Notice what's not on the list: how pretty the website is, how many times it says "quality" and "trusted," or how much you spend on ads. The engines are reading for facts they can check, and they're checking them against third parties.
Where the engines look before they name a contractor
- Your Google Business Profile. Categories (primary and secondary), the services list, hours, service area, photos, and the reviews. For Google's AI Overviews and AI Mode this is the single biggest input; for the other engines it's a major one, because they read Google reviews too.
- Review platforms. Google first; then Yelp, Angi, HomeAdvisor, Thumbtack, the BBB, Nextdoor, Houzz for remodelers, Facebook. Which ones matter in your town is something the prompts will tell you (see below), not something a listings vendor can tell you.
- Licence and insurance records. Where your state or county publishes a contractor licence database, the engines cite it when the homeowner asks "is this company licensed." Make sure the business name on the licence matches the name everywhere else.
- Manufacturer and program credentials. Certified-installer programs (the shingle manufacturers' top-tier programs, factory-authorized HVAC dealer programs, and similar) publish dealer locators. Being listed there is corroboration from a source the engines treat as authoritative, and your site should say you're in the program in the same words the manufacturer uses.
- Your own site. Service pages, price ranges, project pages, the about page, and the schema markup underneath. The engines lift sentences from here when they're specific and skip the page when it's adjectives.
- Local press and community pages. The chamber, the local news story about the storm response, the high school booster page, the Reddit thread where three neighbours recommended you. Small on their own; together they're the evidence that you're real and local.
Run these ten prompts on your own company
Swap in your trade, your town and your competitor. Run each on ChatGPT with search on, Perplexity, Gemini, and a normal Google search to see whether an AI Overview appears. For each answer, note whether you were named, who was, and which sources were cited.
1. Who are the best roofers in [city]? (swap in your trade)
2. My [roof is leaking / furnace stopped / basement is flooding]. Who should I call in [city] right now?
3. How much does [a roof replacement / a new furnace / a bathroom remodel] cost in [city or state]?
4. Is [your company] a good [roofer]? What do reviews say?
5. Compare [your company] and [competitor].
6. Which [roofers] in [county] are licensed and insured, and how can I check?
7. Which [HVAC companies] near [town] offer financing?
8. Who installs [brand or product] in [city]?
9. What should I ask a [contractor] before hiring them?
10. [The question your office answers most often, word for word]The sources column is the whole point. If the engines keep citing Angi and the BBB for your trade in your county and you've never claimed either, you now know where to go. If they cite a competitor's project page with photos and a price, you know what to build. The twenty-minute version of this check is here.
Want to know who the engines name in your town?
Send your website and the competitor you keep losing jobs to. You'll get the ten prompts run across four engines for your trade and service area, who gets named, and the fixes in order. Free, written, within 24 hours.
Prefer to talk? 315.430.0782. I answer my own phone.
The fixes, in order
Cheapest and most-blocking first. The first three are a week's work and they change what every later fix is built on.
1. Answer the phone, including after hours
Not an AI-search item on its face, and it's first for a reason: the emergency prompt ends in a call, and the engines read reviews that say "nobody picked up." A missed-call text-back and an after-hours answer (a person or an honest AI receptionist) turn the lead you've already earned into a job, and the reviews that follow feed everything below. The math on missed calls is here.
2. The fact sheet, and one version of every fact
One page with the exact business name, address, phone, hours, service area, founding year, licence numbers, insurance carrier, certifications, services with price ranges, and the current review counts by platform with the date. Then make the website, the Business Profile, Bing Places, every directory and the licence record match it, word for word. Engines skip companies whose facts disagree with themselves. The template is here.
3. A complete Business Profile
Primary category correct (Roofing contractor, HVAC contractor, Plumber, Electrician, Bathroom remodeler), secondary categories for the services you actually sell, the services list filled in with descriptions, hours that include your real emergency availability, service area drawn where you'll actually drive, and photos of real crews and real jobs, added monthly. Then the review flow: a text with the link the day the job is done, sent by the system, and a reply to every review, especially the bad ones, because the engines read those too.
4. Service pages that state the range
Each service gets its own page with, near the top, two or three sentences a homeowner (or a model) can quote: what the job is, where you do it, what it typically costs and why it varies, what's included, how fast you can start. "Full asphalt roof replacement in Onondaga County typically runs $9,000–$24,000 depending on size, pitch and tear-off; most jobs are done in one to two days; financing available." That sentence gets you named for the cost prompt. "Contact us for a free estimate" gets your competitor named.
5. Project pages with photos, places and numbers
Ten real projects beat a gallery of fifty unlabeled photos. Each page: the town, the problem, what you did, the products used, roughly what it cost, how long it took, and the review that came from it if there is one. These are the pages the engines cite for "who does good work in [town]," and they're the honest version of the town page.
6. Reviews where the engines look
Google always. Then whichever platforms the sources column named for your trade in your area. Claim the profiles, match them to the fact sheet, and point the post-job request at the platform you most need to grow. Volume and recency both matter, which is why the request has to be automatic rather than remembered.
7. Listings and licence records that agree
Google, Bing Places (ChatGPT's search draws on Bing's index), Apple Business Connect, the BBB, the chamber, the manufacturer dealer locators, and the state or county licence database. Same name, same address, same phone. Fix the ones that disagree; don't create twenty new ones to drown them out.
8. Schema markup for the business and the services
A RoofingContractor, HVACBusiness, Plumber, Electrician or GeneralContractor block on the homepage with address, hours, service area, licence and price range; a Service block on each service page with the range; FAQ markup on the questions page. Google says structured data isn't required for its AI features, and it still matters here because it states your facts in a form that can't be misread, and because Bing and the other engines read it. Copy-paste blocks, including the roofer example, are here.
9. A questions page in the homeowner's words
The twenty questions your office answers every week, answered plainly with a specific in each: a price, a timeframe, a brand, a permit requirement. "Do I need a permit to replace my roof in the Town of Clay?" answered with the actual answer is exactly the kind of local, first-hand content Google's guidance says it wants and the AI engines quote.
10. Real town pages, not templates
For the five or six towns where most of your work is, a page with projects from that town, reviews from that town, and anything specific to working there (permit office, common roof types, the HOA that requires architectural shingles). That's a page worth citing. Thirty templated "[trade] in [town]" pages with the name swapped is the pattern Google's spam policies call scaled content abuse, and it's a good way to lose the pages that were working.
How this fits with Local Services Ads and Search ads
If you're in the Local Services Ads unit, you've already done part of this: the licence and insurance check, the Google Guaranteed or Screened badge, the review flow. LSA ranking rewards reviews, responsiveness and proximity; the AI shortlist rewards the same things plus facts it can quote. The difference is the bill. LSA charges per lead (and generously counts the junk unless you dispute it; that math is here). The AI shortlist is free once you're on it, and it keeps producing when the budget pauses.
Search ads stay for the emergency and decided-buyer queries, where an ad reaches a homeowner before any organic result does. What tends to come off the paid budget, once the organic and AI side is working, is the informational spend: "how much does a roof cost," "how long does a furnace last," the queries where your own pages now rank and get quoted. How to find that overlap in your own account is here.
The 30-day plan
| Days | Do | Result |
|---|---|---|
| 1–5 | Run the ten prompts; write the fact sheet; set up missed-call text-back and after-hours answer; turn on the post-job review text | You know who's named and why; no earned lead goes to voicemail |
| 6–12 | Complete the Business Profile; match Bing, Apple, BBB, chamber, dealer locators and the licence record to the fact sheet | One company, one set of facts, corroborated where the engines look |
| 13–22 | Rewrite the first screen of each service page with the range; add the business and Service schema; build the questions page | Something quotable exists for every cost and "who should I call" prompt |
| 23–30 | Publish the first five project pages and the first two real town pages; schedule the monthly prompt check | The compounding part starts, with a way to measure it |
After that: a project page a week, photos to the profile monthly, the prompt log re-run monthly, and a quarterly look at what the paid account is buying that the organic side now wins. For most home-services companies that last check funds everything else on this list.
Questions I get about this
Do homeowners really use ChatGPT to find a contractor?
A growing share do, especially for planned work (a roof, a furnace, a remodel) where they research before calling. Emergencies still go to Google and the phone, and Google's own results now often open with an AI Overview for cost and "who should I hire" questions. You don't need a statistic to act on this: run the ten prompts for your trade and town and see who gets named. If it isn't you, that's the answer.
Does my Google Business Profile matter for AI answers?
More than anything else on this list for Google's AI Overviews and AI Mode, which draw on the same local data as the map pack, and a lot for the other engines, which read Google reviews and the profile's facts. Categories, services, hours, service area, photos and the review flow are the inputs.
How does this fit with Local Services Ads?
They reinforce each other. LSA ranking depends on reviews, responsiveness and a verified licence; so does being named by an engine. The review flow you build for LSA is the same one that gets you corroborated. The difference is that LSA is a per-lead auction and the AI shortlist is free once you're on it.
Should I make a page for every town I serve?
Make pages for the places where you actually have work to show: projects, photos, reviews from that town, the permit or HOA quirks you've dealt with there. Thirty near-identical "roofer in [town]" pages with the town name swapped is the pattern Google's spam policy calls scaled content abuse, and the engines ignore it or worse. Five real town pages beat thirty templates.
We don't publish prices because every job is different. Isn't a range misleading?
A range with its assumptions stated isn't misleading; it's the most useful sentence on your site. "Most full asphalt roof replacements we do in Onondaga County run $9,000 to $24,000 depending on size, pitch and tear-off, as of October 2026" tells the homeowner what to expect and tells the engine what to quote. The competitor who publishes a range is the one the engine names when the homeowner asks about cost.
How fast will this work?
Profile, listing and site-fact fixes show up in search-grounded answers within weeks. Reviews and corroboration build over a quarter or two. The emergency-call fixes (answering the phone, after-hours response) pay back the same week, which is why they're in the first ten days.
