TL;DR: HVAC companies can buy ChatGPT Ads today in OpenAI’s self-serve Ads Manager, with ZIP-code targeting, a $25 daily minimum and a suggested starting bid of $3 to $5 per click. HVAC was one of the first categories Angi tested, per OpenAI’s own case study. The catch is timing: a campaign has no call button, no hour-of-day schedule and no weather trigger, so it is a poor tool for a no-heat call at 6 a.m. It fits the decisions homeowners think over, like repair or replace and heat pump or furnace. Build a cooling and a heating campaign around those decisions before each season, and judge them against what a replacement lead costs you on Google. Our short verdict: worth a test if replacements are a big part of your revenue and you can trace a lead to a sold job; skip it for emergency repair.
This is the HVAC version of our ChatGPT Ads guide for home services, which covers sign-up and a general checklist. Every OpenAI setting below was re-checked in OpenAI’s Ads Manager help center on November 1, 2026. The product is still in beta, so expect changes.
Repair calls vs replacement decisions
A homeowner whose AC quit in July wants a truck today. One whose 15-year-old system just needed its third repair wants to understand options first. ChatGPT Ads serve the second one, because of how the product works:
- No call button. OpenAI’s ads basics page lists what an ad contains: name, logo, title, description, landing page and image. Every lead clicks through to your page first.
- No hour-of-day schedule. A campaign has dates, locations and platforms (campaign guide), and context hints “cannot enforce geographic limits, schedules, or exclusions.” Local Services Ads offer “All day,” business hours or “Custom hours” (Google help).
- Uneven daily spend. The budget is averaged over a Sunday-to-Saturday week, and a campaign that spends its weekly allowance early “may stop delivering until the next budget week.”
| Job | What the homeowner asks first | Fit for ChatGPT Ads |
|---|---|---|
| No heat or no cooling today | “AC repair near me,” then calls whoever answers | Weak. Keep it on Local Services Ads and Search |
| Tune-ups and maintenance plans | Usually nothing; they book the company they used last time | Weak. Low ticket, loyalty driven |
| Repair-or-replace on an older system | “Is it worth fixing a 14-year-old AC?” | Good. A decision, not a reflex |
| Heat pump conversion, furnace or AC replacement | Heat pump vs furnace, sizing, installed price | Best. Big ticket, lots of questions |
| Ductless mini-split for an addition or garage | Whether one works for a single room and what it costs | Best. Planned project |
This sorting is our reading of homeowner behavior, not survey data. The demand is real, though: SE Ranking shows about 5,400 US searches a month each for “heat pump vs furnace” and “hvac replacement cost” (checked November 1, 2026), the same questions people now ask ChatGPT.
Build seasonal campaigns before the rush
OpenAI’s campaign guide tells advertisers to “consider seasonality” when setting dates. For HVAC, that means two campaigns, built weeks ahead, for three reasons:
- You can’t copy last season’s campaign. Per OpenAI, “you can’t clone a Clicks campaign into another Clicks campaign.” The heating campaign is entered by hand, ads included.
- Ads go live after review. OpenAI warns: “don’t publish the campaign just to access these settings. Your ads could start running before you’ve had a chance to set your schedule.” Set the dates, then publish.
- No weather trigger. In a heat wave the only lever is the daily budget, and a day’s spend “will not exceed twice your selected daily budget.” Raise it on the first hot day, not the third.
| Campaign | When we would run it | Ad groups | Leave on Google |
|---|---|---|---|
| Cooling | Late winter to early summer, before the first heat wave fills the schedule | AC replacement, heat pump conversion, mini-splits | No-cooling repair calls |
| Heating | Late summer through fall, before the first cold snap | Furnace replacement, heat pump conversion | No-heat repair calls |
These windows are our suggestion; shift them to your climate. In Phoenix the cooling campaign may run most of the year, in Minneapolis the heating one matters more. Keep them separate either way, as OpenAI recommends “separate campaigns for meaningfully different objectives, regions, or product categories,” so one season’s budget can’t starve the other.
The numbers to beat: what an HVAC lead costs on Google
Most HVAC companies that would test ChatGPT Ads already pay for Google leads, so start there. The most detailed recent HVAC figures we found come from SearchLight Digital, which publishes its clients’ data. In February 2026, HVAC Local Services Ads averaged $51 per lead across 409 accounts, with a 44.0% book rate (leads that became a booked appointment) and a $2,110 average ticket (SearchLight LSA benchmarks). SearchLight notes that “February is heating season,” which helps both numbers. For replacement work, its January 2026 breakdown of non-branded Google Search by service line is more useful (SearchLight HVAC Google Ads benchmarks):
Chart: average cost per lead for US HVAC companies, from SearchLight Digital’s client data. Local Services Ads: February 2026, 409 HVAC accounts. Search rows: non-branded Google Search campaigns by service line, January 2026, from a sample of 816 HVAC and plumbing contractors. Both months fall in heating season. No comparable ChatGPT Ads figure has been published.
| Lead type | Cost per lead | Book rate | Spend per booked appointment (our math) |
|---|---|---|---|
| Local Services Ads, all HVAC (Feb) | $51 | 44.0% | about $116 |
| AC install, non-branded Search | $157 | 24.2% | about $649 |
| Heating install, non-branded Search | $354 | 42.8% | about $827 |
AC install leads were cheap but booked at the lowest rate in the sample, so a booked install appointment cost about $649 in Search spend. That is the bar a ChatGPT replacement campaign has to clear. Caveats: one vendor’s clients, heating-season months, and the Search sample mixes HVAC and plumbing contractors. Google is also moving Local Services Ads into Google Ads as Performance Max campaigns with pay-per-lead goals, starting August 2026 with HVAC among the first categories (Google help), so use your own account’s numbers when you can.
What a ChatGPT click has to do
ChatGPT Ads charge per click (or per 1,000 views), so your cost per lead depends on how many clicks call or fill a form. At OpenAI’s suggested bids:
| To match… | At $3 per click | At $5 per click |
|---|---|---|
| $51 per lead (HVAC Local Services Ads) | 1 lead every 17 clicks (about 6%) | 1 lead every 10 clicks (about 10%) |
| $157 per lead (AC install, Search) | 1 lead every 52 clicks (about 1.9%) | 1 lead every 31 clicks (about 3.2%) |
Beating Local Services Ads on raw cost per lead is a high bar. Beating Search on replacement leads looks reachable with a good landing page, and cost per sold replacement is the comparison that pays the bills.
Ad groups: one per decision, emergency only if you answer 24/7
OpenAI’s ad group guide says to “create separate ad groups when they require different messaging or landing pages.” For HVAC we would build one group per replacement decision:
| What the homeowner asks ChatGPT | Ad group | Landing page |
|---|---|---|
| “My AC is 15 years old and needs a compressor. Repair or replace?” | AC replacement | System options, installed price ranges, install lead time |
| “Heat pump or gas furnace for a cold climate?” | Heat pump conversion | Cold-climate models, dual-fuel option, real installed costs |
| “How much does a new furnace cost?” | Furnace replacement | Prices by efficiency level, next install date |
| “How do I cool a garage conversion without ductwork?” | Ductless mini-splits | Single-zone and multi-zone examples from your jobs |
These prompts are our examples, not OpenAI data. Start with the two groups that carry your best margin for the coming season.
Emergency repair group? Only if someone picks up at every hour, since you can’t limit when the ad shows. A “no cooling” ad that rings to voicemail at 11 p.m. pays for the click and hands the job to the next company.
Try the multi-image format for equipment. Since October 26, 2026, OpenAI offers multi-image ads as an open beta: four to six square images that all link to one page (ad creation guide). Photos of your own installs, a heat pump condenser or a mini-split head on a wall, suit someone comparing options better than manufacturer renders.
Context hints and ad copy for HVAC
Each ad group gets a few context hints: short sentences on what you offer, who it suits and when it helps. OpenAI’s context hints guide shows the level of detail with a plumber: “After-hours residential plumbing in [city], with an $89 service-call fee for urgent repairs.” HVAC versions we would write:
| Ad group | Type | Example context hint |
|---|---|---|
| AC replacement | When | “Homeowners deciding whether to repair or replace a central AC that is more than 12 years old” |
| Heat pump conversion | What | “Cold-climate heat pump installs in Denver and Aurora, with gas furnace backup for the coldest nights” |
| Furnace replacement | Who | “Owners of older homes with a gas furnace past 20 years old and rising repair bills” |
Use your real cities and services. ZIP codes go in the location setting; a hint like “only show this in Denver” does nothing. When a promotion ends, update the hint, as OpenAI asks.
Ad titles and descriptions
OpenAI asks for “multiple title/copy variations per offering” and copy that “explains the practical value of what you offer, who it is for, and when it may be helpful.” Our examples:
| Title | Description |
|---|---|
| Repair or Replace Your AC? Get a Straight Answer | Free replacement estimate in Denver with written options at three price levels. Install usually within a week. |
| Cold-Climate Heat Pumps, Installed in Denver | See which models we install, what they cost and how they handle below-zero nights. |
Watch three claims. The federal heat pump tax credit: the IRS says homeowners “can claim the credit for improvements made through December 31, 2025” (IRS), so a 2026 ad promising it is out of date; name a state or utility rebate only after confirming it is open in your area. “24/7,” only if true at every hour. And financing: OpenAI’s ad policies restrict ads for financial products, so we would keep payment plans on the landing page, not in the ad. Ads must be “truthful and not misleading, including unfounded claims about capabilities, pricing, outcomes.”
The landing page does the selling
The ad can’t place a call, so the page has to. In the ChatGPT phone app the click opens in an in-app browser. What we would put on a replacement page:
- A tap-to-call number at the top, a tracking number used only on ChatGPT pages.
- The decision in the headline, matching the ad: “AC repair or replacement in Denver.”
- Installed price ranges by system type, labeled as typical ranges, with a line that ductwork, electrical work and permits change the final price.
- A short estimate form (system age, home size, ZIP code), your license number and install reviews.
OpenAI asks for “product pages, collections, or content pages rather than defaulting to a generic homepage,” and robots.txt must not block OAI-AdsBot or OAI-SearchBot. Our guide to lead generation landing pages covers the layout.
License numbers in HVAC ads
OpenAI’s policies require licensed advertisers to “maintain those credentials and comply with applicable laws and regulations.” Some states add HVAC ad rules. In Texas, “All advertising by air conditioning and refrigeration contracting companies designed to solicit air conditioning or refrigeration business must include the affiliated licensee’s license number” (16 Texas Administrative Code 75.71). The exemptions cover national TV, phone book listings, phone calls, promotional items, letterhead and shop signs, not online ads. If your state has a rule like this, ask the board how it applies to a short ChatGPT ad, and show the number on every landing page. This is not legal advice.
Count sold replacements, not clicks
Replacement leads take days or weeks to close, so set up tracking before you spend:
- UTM parameters on every ad link; OpenAI says they “persist on ad clicks.”
- A dedicated call tracking number on ChatGPT landing pages.
- The OpenAI Pixel on estimate forms, with the standard “lead_created” event (developer docs).
- Sold jobs sent back if you can, through the Conversions API with the “oppref” click reference (measurement guide). Our guide to offline conversion tracking for contractors covers the same idea for Google and Meta.
- Exclude maintenance plan members. Inclusion audiences need 25,000 matched users, but “exclusions can use smaller audiences.” Upload your plan member list as an exclusion; not every member will match, but each one who does is a click you don’t pay for.
OpenAI and your CRM “may report different conversion totals,” so pick one scorecard: sold jobs and gross profit in the CRM, with a lead source such as “ChatGPT Ads: heat pump” picked by whoever answers the ChatGPT number. The weak spot in most shops is the handoff: the source is set on the call, but the comfort advisor’s estimate and the sold job lose it. Check that the source carries through to the sale in ServiceTitan or whatever CRM you use, or you will never see what a ChatGPT install cost.
A first-season test, judged against your Google numbers
OpenAI has not published HVAC results, and 150 to 250 clicks give a direction, not a verdict. A simple rule: if the calls are real homeowners in your area and estimates are being booked but nothing has closed yet, extend to a spend cap you set in advance. If most calls are out of your area or only want a price over the phone, stop and fix the hints and page before spending more.
Should your HVAC company test ChatGPT Ads this season?
| Test now if… | Wait if… |
|---|---|
| You know what a replacement lead and a sold install cost you on Google | You can’t say what any Google lead costs |
| Replacements, heat pumps or mini-splits are a big share of revenue | Most revenue is repair calls and tune-ups |
| You have a page per system type with real price ranges | Every ad would go to your homepage |
| Calls are tracked by source and sold jobs are in a CRM | Calls come in on one number with no source logged |
On the “wait” side? Fix Google first; our HVAC PPC and marketing services page explains how Leads4Build runs Search and Local Services Ads for heating and cooling companies. Getting named in ChatGPT’s answers for free is a separate job, covered under AI SEO.
The bottom line
For HVAC, ChatGPT Ads are not a no-heat, no-cool channel yet. They are a fair test for replacement decisions homeowners think through. Build a cooling and a heating campaign ahead of each season, give each decision its own ad group and page, and judge cost per sold install against your Google numbers. For the general setup see our ChatGPT Ads guide for home services; if you also run plumbing, see the plumbing version; for B2B, see the B2B version.
Plus one free tool from us, unrelated to ads: we built an AI social media post generator for HVAC companies that writes seasonal, tip and job posts for Facebook and Instagram. It is free and one of 21 trade generators on our tools page.
About the sources: OpenAI settings come from its Ads Manager help center (campaigns, context hints) and ad policies, re-checked November 1, 2026. HVAC cost per lead figures are SearchLight Digital’s published client data (Google Ads benchmarks). Other sources are linked where each fact appears.