What it costs to be left off the AI shortlist, in dollars

Every marketing loss you have ever had showed up somewhere. A lead that did not close is in your CRM. A bad month of ads is in your account. This one leaves no trace at all, because the customer never contacted you. That is what makes it the most expensive kind.

The loss that leaves no trace

A homeowner's AC dies in July. She opens ChatGPT and types "my AC just stopped working, who should I call in my area". She gets an answer naming two companies. She calls the first one. It books the job.

You were never in that conversation. There is no missed call in your phone system, no abandoned form, no lost lead in your CRM, no impression in any ad account. The job happened, the money moved, and nothing in your business recorded it.

You cannot report on a customer who never existed in your systems.

This is why owners underrate it. Every other marketing problem shows up as a bad number. This one shows up as nothing at all.

The four numbers that set the size

The size of the loss is set by four measured figures. None of them are ours.

Figure	Value	Source
US consumers using AI tools for local business recommendations	45%, up from 6% a year earlier	BrightLocal Local Consumer Review Survey 2026, n=1,002 US adults
Consumers who trust AI platforms for business recommendations	40%	BrightLocal 2026
Consumers who look at three or fewer businesses before deciding	72%, with only 1% looking at more than ten	BrightLocal Consumer Search Behavior 2026
Average HVAC ticket	$1,400 service, $7,800 install	PipelineOn, HVAC Leads Guide 2026

Add two more that tighten the window. 75% of consumers decide which business to use in under 30 minutes, and 28% decide in under five. And ChatGPT alone was used for business recommendations by 31% of consumers in the last twelve months.

Put plainly: close to half of buyers now ask a machine, the machine names two or three companies, and most buyers never look past three. The shortlist is not a step in the decision. It is most of the decision.

The math, with your own figures

Here is the formula. Use your own numbers, not ours, because the only version that matters is yours.

Step	What to use
A. Jobs you book in a month from all sources	Your number
B. Share of local buyers now starting with AI	45%, or your own estimate for your market
C. Your Shortlist Share	The share of AI answers for your city that name you. Measure it, do not guess
D. Your gross profit on an install	Your number, not the ticket price

Then: the jobs in play through AI are roughly A multiplied by B. The share of those you are currently eligible for is C. The gap is (1 minus C) multiplied by the jobs in play. Multiply by D for the money.

The one number you cannot invent is C. Which brings us to the honest part: you almost certainly do not know your Shortlist Share, because AI answers change from run to run, so a single screenshot tells you nothing. The method is in how the four AI assistants pick an HVAC company.

A worked example

This is an illustration with made-up company figures, not a measured result and not a promise. Replace every line with your own.

• A company books 60 jobs a month, of which 8 are installs.
• Say a quarter of its new customers now start with an AI assistant. That is 15 jobs a month in play, including 2 installs.
• Its Shortlist Share is measured at 10%, meaning it gets named in one AI answer out of ten for its city.
• So it is absent from 90% of those answers. Roughly 13 jobs a month are being decided in a conversation it is not part of, including close to 2 installs.
• If one install carries $2,500 of gross profit and one service call carries $400, the exposure is in the low thousands per month. Every month. Compounding through a season.

Now the caveat that keeps this honest. Not every one of those jobs would have come to you even if you were named, because you would be one of two or three names and the customer picks one. The useful reading is not "this is the money you lost". It is "this is the size of the pool you are not allowed to compete in".

Compare it to what you already pay

The exposure above is worth comparing against what the same company already spends to buy attention. 2026 HVAC figures, measured on cost per booked job rather than cost per lead:

Channel	Cost per booked HVAC job, 2026
Google Local Service Ads	$168 to $190
Thumbtack	$250 to $260
Angi	$542

Sources: PipelineOn, HVAC Leads Guide: Cost Per Job in 2026, citing SearchLight Digital 2026 (816 contractors, $14.9M spend) and BlueGrid Media 2026. Astra Results Marketing, June 2026. The full comparison is in cost per booked HVAC job in 2026.

An HVAC owner will happily pay Angi $542 for one booked job. The same owner will not spend an afternoon fixing the listings that decide whether a machine names him for free. That is not a judgement, it is just the gap between a bill you receive and a loss nobody invoices you for.

Why this got urgent in one year

Six percent to forty-five percent in twelve months. That is a 7.5-fold increase in the share of consumers using AI tools for local business recommendations (BrightLocal 2026).

A few more markers of how fast the ground moved:

• AI became the third most popular source of business recommendations (BrightLocal 2026).
• AI Overviews now appear in 68% of local searches, against 39% for local packs (Search Engine Journal, 2026).
• Only 37.9% of citations inside Google AI Overviews came from pages ranking in the top 10 for the query, down from 76% (Ahrefs, 2026). Being first is no longer the way in.
• 39% of ChatGPT users click through to a business website when looking for local services (Sagapixel), so these are real visits, not just impressions.
• Consumers over 60, the most cautious group, are already at 24% (BrightLocal 2026).

There is also a seasonal edge that matters in this trade specifically. The questions get asked when the equipment fails, which means the first heat wave and the first cold snap. Being absent in April is cheap. Being absent in July is not.

This is not a bet against Google

Worth saying clearly, because the pitch for AI visibility often oversells itself.

Google did not stop mattering. Google AI Mode cited Google's own pages in 95.5% of home-services answers in the 2026 study, and Google Business Profile alone accounted for 28.5% of local AI citations in Yext's analysis of 6.8 million citations. Nearly everything that gets you named by an assistant, accurate profiles, detailed reviews, clear service and city pages, correct listings, also helps you in Google Maps and ordinary search.

So the work is not a side bet. It is the same hygiene, now graded by a second examiner that nobody was watching until this year.

How to find your own number

You need C, your Shortlist Share. Measuring it honestly takes three rules.

1. Use real customer questions, not keywords. "My AC is blowing warm air, who do I call in [town]" is what people type. "Best HVAC [town]" is what marketers type.
2. Repeat every question several times on each assistant. Answers are non-deterministic. One run is an anecdote. Report the share with the sample size, like "named in 3 of 15 answers".
3. Track the trend, not the run. Week over week movement is the only signal that survives the noise.

Then write down who gets named instead of you, and which sources those answers cited. That list is your work order. Nobody can guarantee what an AI will say, and anyone who promises you a position is selling something that does not exist. What you can do is fix the signals it reads and watch the line move.

Questions owners ask

How do I know if AI is costing my HVAC company jobs?

You cannot see it in your own systems, because the customer never contacted you. The only way is to measure directly: ask real customer questions for your city on each assistant, several runs each, and record how often you are named versus your competitors.

What is Shortlist Share?

The percentage of AI answers for your trade and your city that name your company. Because AI answers change from run to run, it only means something when measured across repeated runs with the sample size shown, for example named in 3 of 15 answers.

Is 45% of consumers using AI really credible?

It comes from BrightLocal's Local Consumer Review Survey 2026, run on a representative panel of 1,002 US adults, which recorded 45% using ChatGPT or another generative AI tool for local business recommendations, up from 6% in the prior year's survey. ChatGPT alone was used by 31%.

If I fix this, will I definitely get more jobs?

Nobody can promise that, and an AI answer names only two or three companies even when you are on it. What the work does is make you eligible for a pool of decisions you are currently excluded from, and almost all of it also improves Google Maps and ordinary search.

Should I stop spending on Google Ads and lead services?

Not on the strength of this. Google Local Service Ads were the cheapest measured HVAC channel in 2026 at roughly $168 to $190 per booked job. The point of comparison is that owners pay $542 per booked job on Angi while leaving free citation work undone.