It is 9:40 on a Tuesday night. The AC quit in a house full of kids. The homeowner does not open Google and scroll ten blue links. She opens ChatGPT and types, “Who's the best AC repair company near me that's open right now?”
In that moment, your website did not get a click. It got an interview. ChatGPT, Perplexity, and Gemini asked a series of questions about your business, checked the answers against a dozen different sources, and picked a name. Maybe it was yours. Increasingly, it is not.
The AI interview is the verification process answer engines run before they recommend a business. The engine asks: Who are you? What exactly do you do? Where do you work? Can I verify you anywhere besides your own website? Have customers vouched for you lately? Can I extract a clean, confident answer about you? Only the businesses that pass get named. Everyone else is invisible, and no amount of ad spend fixes that.
Your rankings will not save you
A 2026 SOCi local visibility index, covered by Search Engine Land, measured how often business locations get recommended by AI assistants versus how often they show up in Google's local results. About 1.2% of locations were recommended by ChatGPT, 11% by Gemini, and 7.4% by Perplexity, compared with 35.9% appearing in Google's local 3-pack. Their estimate: earning AI visibility is three to 30 times harder than ranking well in traditional local search.
You can own the map pack and still lose the customer, because she never saw the map pack. She saw one answer, and your competitor's name was in it.
Fewer than half of the brands that lead in Google local visibility also lead in AI results. Two different games, two different scoreboards, and most contractors are still training for the old one.
Rankings are a participation trophy. Recommendations are the game.
What the answer engine actually asks
Forget keywords for a minute. Think like the machine. When a homeowner asks for a recommendation, the engine runs a background check on your business before it says your name out loud. Here are the six questions it asks, in order.
1. Who are you, exactly?
This is entity branding, and it is the foundation everything else sits on. The engine needs to know you are one specific, real business, not three similar-sounding companies sharing a phone number.
Pick your exact business name and use it identically everywhere. Not “Bob's HVAC” on Yelp, “Bob's Heating & Air Conditioning LLC” on your website, and “Bobs HVAC and Plumbing” on Facebook. One name. Same spelling, same punctuation, same suffix, every place your business appears. Same goes for your categories and your services list. Your Google Business Profile is the source of truth, and everything else should mirror it, not approximate it.
Inconsistent identity is the fastest way to fail the interview. When the engine finds three versions of you, it does not pick the best one. It picks someone else.
2. Where do you actually work?
State your real service area plainly. List the cities and neighborhoods you genuinely serve. If you are a service-area business with no customer-facing storefront, say so on your profile instead of hiding it.
Do not publish fifty near-identical location pages for towns you have never worked in. Answer engines are pattern-recognition systems. They spot manufactured geography, and it reads as untrustworthy. One honest service area beats a fake empire every time.
3. Can I verify you anywhere besides your website?
Your website is your testimony. Everything else is corroboration. Answer engines synthesize from many sources at once: your site, directories, reviews, news, social platforms, forums, podcasts. A business that only exists in one or two of those places barely registers.
The sources that carry the most weight for trades businesses are the boring ones. Google Business Profile. Yelp. BBB. Angi and HomeAdvisor. Nextdoor. Facebook. Your local Chamber of Commerce. Industry directories. The engine wants the same facts repeated by independent sources: name, address, phone, hours, services. When ten independent sources agree, you are verifiable. When they contradict each other, you are a risk, and the engine moves on.
You cannot outrank bad data. Fix the data.
4. Have real customers vouched for you lately?
Reviews are not testimonials anymore. They are testimony, and the answer engine treats them that way.
Two things matter. First, recency and velocity: a shop with 300 reviews that stopped getting them in 2024 looks closed to a machine, while a shop with 120 reviews and six new ones this month looks alive. Second, sentiment and substance: AI recommendations consistently favor above-average ratings, and reviews naming specific services, techs, and neighborhoods give the engine extractable facts.
The playbook is unglamorous: ask every customer with a direct link or QR code on the invoice, and respond to every review within 48 hours. A thoughtful reply to a one-star review is worth more than ten unanswered five-star reviews, because it proves a human runs the place.
5. Can I pull a clean answer from your content?
When Perplexity or ChatGPT cites a source, it lifts a passage. The passages that get lifted are short, definitive, and entity-rich. That is a quote from the research, not a guess, and it tells you exactly how to write.
Stop writing pages for a keyword tool and start writing answers for a homeowner. Build a real FAQ that answers the exact questions people ask, in the words they use: “How much does it cost to replace an AC unit in Houston?” “Why does my breaker keep tripping when the AC kicks on?” Put the direct answer in the first sentence. Then explain. Then add the local detail the engine cannot get anywhere else: permit rules, climate-specific problems, realistic price ranges for your market.
Then make it machine-readable: FAQ schema on your Q&A pages, LocalBusiness schema with your name, address, phone, hours, and service area marked up explicitly. Structured data hands the engine your facts on a platter instead of making it guess.
6. Are you safe to recommend?
This is the question nobody talks about, and it decides close calls. When the engine names you, it puts its credibility on the line. It looks for proof you are legitimate and accountable: license numbers on your site, real photos of your techs and trucks, a physical address, guarantees in plain language, an about page with actual humans.
Stock photos and vague “satisfaction guaranteed” copy fail this check. A license number, a real crew photo, and a written guarantee pass it. Trust is not a vibe. It is evidence.
How to check what AI currently says about you
You would not run ads without checking the landing page. Do not run a business in 2026 without checking what the machines say about it. Twenty minutes:
- Open ChatGPT, Perplexity, and Gemini in separate tabs.
- Type the exact question a homeowner would ask, for your top three services and your top three cities. “Best plumber in Katy for a slab leak” beats “plumber Katy.”
- Screenshot every answer. Save them in a folder with today's date.
- Note who gets named, who gets skipped, and which facts about your business are wrong. Profile data is only about 68% accurate on ChatGPT and Perplexity, per the same SOCi index, so expect errors.
- Ask a follow-up: “Why did you recommend them?” The engine will often tell you which signals it used. Free competitive intelligence.
- Trace every error to its source. Wrong hours on ChatGPT usually means wrong hours on a directory the engine trusts. Fix the source, not the chatbot. You cannot argue with the machine. You can only feed it better facts.
- Repeat monthly. This is a vital sign now, like checking your reviews.
The 30-day pass-the-interview playbook
Week one, lock the identity. Write down your canonical business name, address, phone, hours, categories, and service list exactly as they appear on your Google Business Profile. This document is now law: nothing gets published anywhere unless it matches.
Week two, fix the top ten sources. Google Business Profile first, then Yelp, BBB, Facebook, Angi, HomeAdvisor, Nextdoor, Apple Maps, Bing Places, and your Chamber listing. Match the canonical identity exactly, fill every field, upload real photos. Tedious, and the highest-leverage work on this list.
Week three, start the review machine. Build the ask into closeout: tech finishes the job, sends the direct review link before leaving the driveway. One person responds to every review within 48 hours. Track reviews per week like you track calls booked.
Week four, publish the answers. Write five FAQ pages for your five most-asked homeowner questions: direct answer first, local detail included, FAQ schema installed. Add LocalBusiness schema to your homepage and contact page. Put your license number and guarantee on the site in plain text, not buried in an image.
Ongoing, measure mentions, not just rankings. Monthly, run the three-engine audit. Track whether you get named, how often, with what facts. When the answers about you get more accurate, you are winning. That is the new scoreboard.
The shops that win the next five years will not be the ones with the prettiest websites. They will be the ones the machines trust enough to recommend. The interview is already happening, with or without you. Pass it on purpose.
Watch the companion conversation
The AI Search Revolution: How Home Service Businesses Win in the Age of ChatGPT — the companion conversation to this playbook, breaking down exactly how home service businesses win when the search box becomes a conversation.
