Key Takeaways
- AI assistants do not rank pages, they pick businesses. Your HVAC page wins by handing the model clean, repeated facts about who you are and where you work, not by stuffing keywords.
- Gemini and Google AI Overviews lean on your Business Profile and reviews. ChatGPT leans on Foursquare. Perplexity leans on comparison content. The foundation is shared, the depth is per platform.
- Build one real page per core service and one real page per city you actually serve. Skip the fifty templated doorway pages with the city name swapped. They get ignored.
- Add HVACBusiness or LocalBusiness schema plus a Service node per page, and validate the rendered HTML. A page can look perfect in the editor and ship with no schema at all.
- DIY foundation work is $0 in software and 4 to 6 hours. Done-for-you AI search retainers for HVAC run $800 to $3,000 per month depending on how many cities you serve.
- Expect Gemini and AI Overviews movement in 2 to 4 weeks, ChatGPT and Perplexity in 6 to 12 weeks. Build seasonal pages ahead of the season.
A homeowner with a dead furnace at 6am does not open Google and scroll through ten links anymore. They ask their phone for "an HVAC company near me that does emergency furnace repair" and ChatGPT or Gemini names two or three companies. If you are not one of them, you do not exist for that customer. It does not matter that you rank number one on Google in your town.
That is the shift. Most HVAC pages are still built for the old game. They are built to rank, not to get recommended by a machine that reads facts, not marketing copy. Here is how to fix that, page by page.
Why Near Me Searches Changed
The phrase "near me" used to mean Google Maps. You optimized your Business Profile, collected reviews, and showed up in the three-pack. That still matters. But a growing share of those searches now happen inside an AI assistant, and the assistant does not show a map. It gives an answer.
When ChatGPT answers "best HVAC company near me," it is not running your old keyword rankings through a ranker. It is assembling an answer from structured data, directory listings, reviews, and web context it has indexed. The companies it names are the ones with the cleanest, most consistent, most machine-readable presence. That is a different target than a page-one ranking, which is exactly why so many well-ranked HVAC companies get zero AI citations. We dug into that gap in detail in why your page-1 Google rankings do not get you cited by AI search. The other shift: each platform pulls from different places. Gemini sits on Google data, ChatGPT pulls hard from Foursquare, Perplexity rewards comparison-style web content. So "optimizing for AI" is not one job. It is a shared foundation plus three flavors of depth.
The HVAC Page Structure That Gets Cited
Forget the service page that opens with "Welcome to our family-owned business serving the community for over 20 years." A machine cannot do anything with that. Here is the structure that gets pulled into answers.
Lead with the direct answer. If the page is about emergency AC repair in a specific city, the first lines should state exactly that: what you fix, where, and how fast you respond. AI assistants quote the part of the page that answers cleanly. Bury that under a mission statement and you lose the citation.
Name the symptoms and the jobs. Real customers describe problems, not service categories. "AC blowing warm air," "furnace short cycling," "thermostat not responding," "ice on the outdoor unit." List the actual symptoms and tie each to the service that fixes it. This is the language people type into assistants, and it is the language the model matches against.
Include a short FAQ on the page itself. Question-and-answer format is the single most quotable structure for AI. "How much does a furnace replacement cost in [city]?" with a specific dollar range gets pulled verbatim. Vague answers get skipped. This is the same dynamic we covered in why ChatGPT does not recommend your local business even with great reviews: the model rewards specificity, not polish.
State your service area in plain text, not just on a map widget. Write the cities and neighborhoods you serve in the body copy, because the assistant reads text far more reliably than an embedded map.
Schema and Data Signals That Move the Needle
This is where most HVAC pages fail silently. Use HVACBusiness or LocalBusiness schema with every field filled: legal name, address, phone, geo coordinates, areaServed, hours, and aggregateRating. On each service page, add a Service schema node tied to the business with the specific service name and the cities it covers. Add FAQPage schema for the on-page questions.
The reason this matters: you are handing the model exact facts instead of making it infer them from prose. A model with clean structured data about your phone, service area, and rating is far more likely to surface you than one piecing it together from a paragraph.
One warning we hit repeatedly: validate the rendered HTML, not the source. Some site builders strip JSON-LD before the page ships, so it looks correct in your editor and arrives at the crawler with nothing. Pull up the live page, view source, and confirm the schema is there. For a system to confirm the AI engines see and cite you, we walk through it in how to track if AI engines are citing your business.
Service Pages vs City Pages
You need both, and they do different jobs. Do not confuse or duplicate them.
Service pages go deep on the work. An AC installation page explains sizing, the difference between repair and replacement, what a typical install runs, and the symptoms that mean it is time to replace rather than fix. These pages answer "how much does X cost" and "do I need X or Y," which are exactly the decision-stage questions assistants get asked.
City pages go deep on the place. A page for a specific metro names the neighborhoods you cover, references real local conditions (hard water shortening water heater life, coastal humidity loading up AC systems, undersized ductwork in older housing), and scopes your services to that city. The local specifics separate a real city page from a doorway page.
The trap: spinning up fifty city pages by find-and-replacing the city name. AI assistants ignore thin templated pages, and Google has penalized them for years. One genuinely local page per city you actually serve beats fifty hollow ones. Serve six cities, build six strong pages, not sixty thin ones.
The Citation Network HVAC Companies Ignore
Your pages do not exist in isolation. AI assistants cross-reference your business against directories and citation sources, and HVAC companies almost always over-index on Google.
ChatGPT pulls heavily from Foursquare. If you have 300 Google reviews and a blank Foursquare profile, you are invisible to a large share of ChatGPT recommendations. Claim Foursquare, Bing Places, and Apple Maps with identical name, address, and phone. Then add the HVAC-specific signals: industry association memberships, manufacturer dealer listings (Carrier, Trane, Lennox locators), and chamber listings. These authoritative citations tip an assistant from "maybe" to "recommend." Perplexity in particular rewards this cross-referenced authority, which we break down in how to get your local business cited by Perplexity AI in 2026.
What This Costs
DIY foundation work costs $0 in software and 4 to 6 hours: claim Foursquare, Bing Places, and Apple Maps, fix NAP consistency, add LocalBusiness and Service schema, and tighten your Business Profile. Any HVAC owner can do this in a weekend.
A done-for-you AI search retainer for a local HVAC company runs $800 to $3,000 per month. The lower end covers foundation plus monthly maintenance for a single metro. The higher end covers multi-city expansion, ongoing page production, citation network building, and per-platform tracking. If you serve one city and have time, DIY the foundation and decide later whether to hire for depth. If you serve five metros and the phone needs to ring in all of them, the retainer pays for the production volume you cannot do yourself. Avoid the $99-per-month offers; that is a directory-submission tool wearing a marketing label.
Your First Week Plan
Day one, ninety minutes: claim Foursquare, Bing Places, and Apple Maps. Confirm your name, address, and phone are identical everywhere. Add or fix LocalBusiness schema on your homepage.
Day two, sixty minutes: rebuild your top service page using the structure above. Direct answer first, symptoms named, on-page FAQ with real dollar ranges, service area in plain text, Service plus FAQPage schema. Validate the rendered HTML.
Day three, sixty minutes: run baseline queries. Ask ChatGPT, Gemini, and Perplexity for an HVAC company in each city you serve. Write down where you appear, where you do not, and which platform shows the biggest gap. That gap tells you where to spend next.
Rest of the week: build one real city page for your strongest metro, add three to five authoritative citations (associations, dealer locators, chamber), and start a Foursquare review push, the one HVAC companies always neglect.
That is the whole foundation. It is not complicated, just specific, and specificity is exactly what most HVAC pages lack. If you would rather have someone build the page system and citation network for you, that is the work we do every day. Book a call and we will map your AI search gaps across ChatGPT, Gemini, and Perplexity, then show you exactly which pages to fix first.
Want to see whether it is worth hiring out versus handling in-house first? Read when to DIY AI search optimization vs hire an agency and then grab a slot to get a custom plan for your metro.
About the author
Matthew Johnson is the founder of Pleiades Consultancy. He previously scaled his own marketing agency to multiple six figures before serving as CMO of an Amazon agency, where the client base tripled from 15 to 45 active clients during his tenure. He worked with some of the largest names in e-commerce, including Ridge Wallet, HexClad, BK Beauty, The Woobles, Walkize, Lonely Planet, and Obvi. He now works with local businesses to maximize their client acquisition and visibility through AI search with ChatGPT, Claude, Gemini, Perplexity, and Bing Copilot.