AI Marketing for HVAC Companies

Something has shifted in how HVAC replacement sales begin, and most contractors are experiencing it without naming it. Homeowners now arrive at the estimate having already asked an AI tool whether they should repair or replace, what a system should cost, and whether they qualify for a rebate. They arrive with a number in their head and a position they are defending. When those answers are good, this helps you, because the customer is informed and the conversation is faster. When they are wrong, and on rebates and pricing they frequently are, your technician spends the visit arguing with a chatbot rather than selling. Reported figures also show AI Overviews now appearing on a large share of search queries and reducing paid click-through substantially, which means this is a traffic question as well as a sales one.

1The Repair or Replace Answer

This is the question these tools get asked most in heating and cooling, and the answer they give is usually a rule of thumb applied without any of the facts that determine the outcome.

• Generic formulas dominate the answers. Age multiplied by repair cost against some threshold, which is a heuristic rather than an analysis.
• The variables that matter are absent. The specific failure and whether it recurs, refrigerant considerations, ductwork condition, efficiency difference, remaining warranty, and how long the owner is staying.
• Answers skew toward replacement. Which sounds like it favours you and does not, because a homeowner primed to expect replacement distrusts the technician who confirms it.
• Publish the real analysis. What actually determines the answer, with the cases where repair is correct stated plainly.
• These tools favour content that distinguishes. Nuanced material that gives conditions rather than a single rule tends to be drawn on more.
• Being the cited source frames the conversation in your favour. The homeowner arrives having read your reasoning rather than a national aggregator's.
• Do not publish a rule of thumb yourself. It is what gets extracted, and it will be quoted back at your technician.

2Where Pricing Answers Go Wrong

• National averages ignore local labour and permitting. A figure that is reasonable in one market is far off in another, and homeowners do not know that.
• Scope is invisible in a chat description. Ductwork modifications, electrical work, code-required upgrades, and disposal all move a price and none appear in the question.
• Equipment tiers get collapsed. A range spanning basic to premium equipment is quoted as though it were one number.
• Publish local ranges with the variables named. Which is more useful than the average and positions you as the accurate source.
• Explain what moves a quote. The same content that wins comparison shopping also corrects the AI answer.
• Anticipate the objection in your sales process. Technicians should expect a number from a chatbot and have a calm explanation ready rather than a defensive one.
• Keep your published ranges current. Equipment costs move, and a stale range on your own site becomes the wrong answer these tools repeat.

3Rebates Are the Worst Answered Topic

Of everything homeowners ask about heating and cooling, incentives are where these tools are least reliable, because programs change faster than training data and vary by state, utility, and equipment.

• Expired programs get quoted as current. Which produces homeowners expecting money that no longer exists, and that disappointment lands on your technician.
• Local utility programs are usually missing entirely. These are frequently the largest incentives available and the least represented in general answers.
• Eligibility conditions get dropped. Efficiency thresholds, equipment types, and documentation requirements disappear from the summary.
• Maintained local incentive pages are the corrective. Dated, specific to your utilities and state, and reviewed on a schedule.
• Name the utility and jurisdiction explicitly. So extraction cannot strip the locality from the answer.
• Frame amounts as current program terms. Subject to change, with a pointer to the program itself, since your page should not be the source of anyone's tax position.
• Check monthly what the tools claim about your area. Both to know your visibility and to brief your sales team on what customers believe.

4Sizing and Equipment Selection

• Square footage rules of thumb are common and wrong. These tools frequently offer a tons-per-square-foot answer that ignores insulation, orientation, windows, infiltration, and ductwork.
• Homeowners then resist a calculated load. Arriving convinced they need a specific size and treating your calculation as an upsell or a downsell.
• Publish why calculation beats estimation. And what oversizing actually costs in comfort, humidity control, and equipment life.
• Cover heat pump questions specifically. Cold weather performance and dual fuel are the areas where generic answers are most outdated relative to current equipment.
• Address brand comparison questions. Homeowners ask these tools which brand is best and get answers assembled from reviews rather than from installation reality.
• Explain that installation quality outweighs brand. Which is true, underexplained, and repositions the entire comparison in favour of a good contractor.
• Do not publish a sizing calculator. It has the same problem as the AI answer and puts your name on a number you did not verify.

5AI Overviews and Your Traffic

• The traffic effect is measurable. Reported analysis puts AI Overviews on a large share of tracked queries with substantial reductions in paid click-through on those queries.
• Informational HVAC content is the most exposed. How systems work, what a component does, and general maintenance advice all summarize cleanly.
• Decision and local content resists it. Repair or replace with local costs, and rebate content tied to a specific utility, are harder to generalize.
• Emergency and booking queries are least affected. Someone with no heat is not reading a summary.
• Weight investment accordingly. Toward decision content, local content, and service pages rather than general explainers.
• Being cited still carries value. Name exposure in a credible answer reaches someone who may search you directly.
• Measure informational content on assists. Not sessions, which will fall regardless of quality.

6Being Named When They Want a Contractor

• Trades get named more readily than regulated professions. These tools will surface local contractors, which is an advantage worth working.
• Dealer locators are strong corroborating sources. Manufacturer authorized listings are exactly the kind of verifiable third-party record these systems reach for.
• Reviews and profile completeness feed recommendations. The same assets driving local rankings.
• Specific queries name businesses more often. A brand plus a service plus a city produces named results far more readily than a general request.
• Certification and dealer tier narrow the field. Which is what makes a small identifiable set these tools can name.
• Aggregators still occupy much of this space. The same competitors that hold organic results.
• Pre-hiring questions are targetable. What to ask an HVAC contractor is a query where a good page gets cited.

7Crawler Access

• Check robots.txt for the major agents. GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and Applebot-Extended.
• Contractor site platforms often set defaults you did not choose. Worth checking rather than assuming.
• Check the CDN and security layer too. Edge rules override robots.txt entirely.
• Google-Extended is separate from search indexing. Blocking it affects AI features only.
• Recheck after site work. Rebuilds reintroduce blocks routinely.
• Permitting is almost always correct here. There is nothing on a contractor marketing site worth withholding from these systems.

8Content Structured for Extraction

• Question headings in homeowner language. Situations and decisions rather than equipment terminology, answered immediately below.
• Name your market and utilities throughout. The strongest signal available on locally variable topics like pricing and rebates.
• Give conditional answers, not single rules. If the failure is this and the age is that, then this, which is both more accurate and more citable.
• Answer cost questions directly. With ranges and variables, since these tools surface cost answers constantly and prefer sources that provide them.
• Byline to a named person with credentials. Role and years in the trade on substantive content.
• Visible FAQ sections with matching schema. The questions your call centre and technicians field.
• Keep answers self-contained. A paragraph that stands alone when extracted is what gets used.
• Date and review on the incentive schedule. Quarterly at minimum on anything involving programs or pricing.

9Prompt Audits and Tracking

• Build a prompt set across the buying journey. Repair or replace scenarios, cost questions, rebate eligibility, sizing, brand comparison, and contractor recommendations.
• Run monthly across the major tools. ChatGPT, Perplexity, Gemini, and Google AI Overviews.
• Check rebate accuracy first. The topic with the highest error rate and the most direct effect on your sales conversations.
• Test pricing prompts for your market. So you know what number customers arrive with.
• Brief the sales team on findings. This audit is genuinely a sales enablement exercise as much as a marketing one.
• Record who gets named. Your business, competitors, aggregators, or nobody.
• Track trends across runs. Rather than treating a single response as definitive.

10Measuring AI Visibility

• Citation rate across the prompt set. Month over month.
• Accuracy rate on local rebates and pricing. What customers believe before your technician arrives.
• Referral traffic from AI tools. Modest volume, high intent.
• Branded search growth. The compounding indicator.
• Ask at booking. Add an AI option to how they found you and record what they had already been told.
• Customer expectations logged at the estimate. Price expectations and rebate assumptions, which tell you exactly what to publish next.
• Estimate conversations that started with a chatbot number. A real and increasingly common sales metric.
• Keep expectations proportionate. A position being established rather than a channel booking jobs today.