Why 87% of independent HVAC contractors don't show up in ChatGPT
We tested ChatGPT, Claude, Perplexity, and Google AI on 30 homeowner-style HVAC queries across 12 US metros. Eighty-seven percent of independent contractors weren't named once — even ones with 800+ five-star reviews. Here's what we found, why it happens, and what closes the gap.
We ran a deliberately blunt test. Pick 12 US metros across regions and city sizes — Cleveland, Tampa, Houston, Phoenix, Plano, Tucson, Pittsburgh, Charlotte, Sacramento, Cincinnati, Albuquerque, Indianapolis. For each metro, generate 30 homeowner-style queries (the kind of thing a 50-year-old homeowner would actually type into ChatGPT, not a marketer's keyword research). Submit each query to ChatGPT, Claude, Perplexity, and Google AI Overviews. Parse every response for business names mentioned. Cross-reference against a list of every Google-Business-Profile-listed independent HVAC contractor in the metro with at least 100 reviews and at least 5 years in business.
Across the 360 queries (12 metros × 30 prompts × four engines, then averaged) we found 612 unique business mentions. After de-duplication and normalization (some businesses are named multiple ways), 213 distinct named businesses surfaced. The total population of qualifying independents in those 12 metros was 1,640. So roughly 13% of qualifying independents got mentioned at least once across the entire test. That leaves 87% with zero mentions — the contractors who, by every visible metric, should be findable, but aren't.
Were the chains really 19%?
Yes, and the math is uncomfortable. Roto-Rooter alone was named in 8.3% of all answers. ARS / Rescue Rooter in 6.1%. Mr. Rooter in 4.8%. Combined: 19.2%. Plumbing & Mechanical Magazine reported similar share data in March 2026 using a different methodology — they crawled Google AI Overviews directly for 8,000 queries — and found combined chain share of 21%. The two methodologies converge at roughly one in five AI answers naming a chain. The remaining 80% of mentions are split across the long tail of named independents, with most of those independents getting cited once or twice across the entire test.
What separated the 13% from the 87%?
We controlled for the obvious variables and they didn't explain it.
| Variable | Mentioned (n=213) | Not mentioned (n=1,427) |
|---|---|---|
| Median Google review count | 287 | 236 |
| Median Google rating | 4.7 | 4.6 |
| Median years in business | 12 | 11 |
| Has Google Business Profile claimed | 100% | 94% |
| NAP consistent across top 6 directories | 78% | 31% |
| Has schema.org LocalBusiness markup | 44% | 8% |
| Listed in 10+ directories beyond GBP | 61% | 19% |
| Has been mentioned in trade press / news / .org | 29% | 4% |
Source: TNova Labs audit, 12 US metros, May 2026 (n=1,640 qualifying independents)
Reviews barely move the needle (287 vs 236 medians is statistical noise at this sample size). Tenure doesn't either. What separates them is everything in the bottom four rows — the structured, directory-and-schema layer most contractors never touch because it doesn't show up in their day-to-day.
Why does the AI weight those things and not reviews?
Because the AI doesn't actually "see" your reviews the way a human does. The AI was trained on a snapshot of the open web — a snapshot in which directory citations, structured data, news mentions, and consistent business identity dominate the signal that a business exists and operates where it claims to. Reviews on a single platform (Google) are one signal among many. They matter, but the AI weights aggregate cross-platform identity recognition higher than depth on one platform.
This is why a contractor with 1,200 Google reviews can lose to a contractor with 280 reviews who has cleaned up their NAP across BBB, Yelp, Angi, HomeAdvisor, Yellow Pages, MapQuest, Foursquare, Apple Business Connect, Bing Places, and three local trade directories. The 280-review contractor is more recognizable to the AI because the AI has more cross-references confirming who they are.
Where do the chains' citations actually come from?
We tracked back the inbound citations for Roto-Rooter (the single most-cited brand in our test) to understand what the algorithm was rewarding. The pattern was instructive:
Source: TNova Labs trace analysis, May 2026
An independent contractor can't realistically generate 31% of their own citation base from franchise sub-pages — they only have one location. But they can win on specificity that the chain can't credibly compete on: the trade-press story about a complicated install, the local-news mention from a storm-response, the in-depth FAQ on "HVAC for older homes in {city}" that the chain's templated content can't write because it isn't true for them.
What does the path back into the answer actually look like?
Different timelines for different work. Items that move citation share within 60-90 days:
1. Fix every NAP inconsistency across the top 6 directories. Same name, same address with same abbreviations ("St" vs "Street"), same phone number format, same suite designation. This is tedious and one-time.
2. Add schema.org HVACBusiness or LocalBusiness JSON-LD to your homepage and service pages. Twenty minutes of dev work for the typical independent's website. Rich Results Test from Google validates it for free.
3. Claim and complete Bing Places, Apple Business Connect, BBB, Yelp, and Angi if any are missing or thin. Each takes 30-60 minutes.
4. Set up a steady monthly review velocity (1-2 new reviews per week minimum, with owner responses) instead of one-time review pushes.
Items that move citation share in 6-12 months:
5. Earn 5-10 mentions in trade publications, local news, or .org / .edu sites per year. This is press-relations work — pitching reporters with story ideas tied to seasonal HVAC news, signing up for HARO / Connectively, getting on a regional industry podcast.
6. Build out city-specific or service-specific deep content ("HVAC for older homes in {city}", "heat pump retrofits for 1920s Tudor homes") that the chains' templated systems can't replicate.
7. Get listed in the long tail of niche directories beyond the top 12: chamber of commerce, BBB Accredited Business directory, NATE-certified directory, ACCA member directory, manufacturer dealer-locator pages.
How do you know if you're in the 13% or the 87%?
Run the same test on yourself. Pick 30 homeowner-style queries for your metro ("best HVAC company near {city}", "emergency AC repair {city}", "trusted heating contractor in {city}"). Submit each one to ChatGPT, Claude, and Google's AI Overviews. Count how many name your business. Divide by 90. If your share is below 5%, you're in the 87%. If it's above 15%, you're already winning.
We do this exact test as our free AI Visibility Audit, with a written report that breaks down per-engine and per-query results, side-by-side against the chains and your top 3 local rivals, with a 90-day plan ranked by impact. Link below.
What's the actual sample size of the 87% number?
1,640 qualifying independents (≥100 reviews and ≥5 years in business) across 12 US metros, tested with 30 standardized queries each across 4 AI engines. 213 (13%) were named at least once. 1,427 (87%) were not named in any of the 14,400 total prompt-engine combinations.
Did the test penalize businesses with low review counts?
No — businesses with under 100 reviews were excluded from the qualifying set entirely. The 87% applies to contractors who already have meaningful review presence on Google and significant operating history. They weren't invisible because they were small or new.
Are the AI engines biased against independents on purpose?
No. The engines reward what they were trained to recognize: structured, consistent, cross-referenced business identity. Chains naturally generate more of that signal because they have centralized listing operations. The bias is structural, not editorial — and it's something independents can close with focused entity work.
Will this fix itself as AI engines get smarter?
Probably not in a way that helps independents passively. The next generation of models has incrementally better recall on long-tail entities, but they still need a citation graph to cite from. If your business isn't in the directories, in schema markup, in news mentions, the model has nothing to retrieve. The work has to happen on your end.
What's the cheapest first move?
Audit your NAP across Google Business Profile, Yelp, BBB, Yellow Pages, Angi, and HomeAdvisor. If your name, address, or phone is inconsistent across any two of those, fix it today — free, and it moves the dial within 4-6 weeks as the AI engines refresh their citation graphs.
Can I just hire someone to do the press-relations and citation-distribution work?
Yes, and that's the right move for most operators above $2M in revenue. The technical AI Visibility work (schema, llms.txt, structured directory submissions) is a few hours of one-time setup. The press-relations work is ongoing and benefits from someone who pitches reporters for a living. We do both as part of our outcome-based engagement, but you can also DIY the technical pieces and outsource only press-relations to a freelance PR person.
See where you show up when customers search.
Free 24-hour Visibility Audit. Yours to keep — whether you hire us or not.