When homeowners ask AI assistants for HVAC recommendations, the answer is almost never a local contractor,it's Angi, HomeAdvisor, or a national franchise. This isn't because your website lacks proper technical setup or schema markup. The real reason is structural: AI models assemble answers from a third-party corpus of data sources, and aggregator platforms dominate that corpus far more than individual businesses do.

Why AI Models Prefer Aggregators Over Local Businesses?

AI doesn't read your website and judge your work the way a human customer might. Instead, it pulls information from a distributed set of sources including your Google Business Profile, review platforms like Yelp and Trustpilot, the Better Business Bureau, local directories, chamber pages, Reddit threads, and industry listings. Your own website ranks as only one input among several, and rarely the most authoritative one.

Aggregators like Angi win because they solve a fundamental machine problem: they package hundreds of businesses, their ratings, and side-by-side comparison data into a single structured, machine-readable document. A retrieval system can parse and cite that far more easily than one contractor's marketing copy scattered across their homepage. This creates what researchers call a "corpus-share gap." You may do better work than any company Angi lists and still lose, because the web has written far more about Angi than about you.

One marketing manager who systematically tested ChatGPT recommendations over a month captured the frustration: "Competitors show up consistently, we barely appear despite stronger traditional SEO. Reverse engineered what they have that we don't... heavier forum presence, third party blog mentions, almost nothing on their own site that we don't also have." The takeaway is blunt: being good isn't the lever. Corpus share is.

What Review Signals Actually Matter to AI Systems?

Before naming anyone, AI models weigh review and trust signals across sources. They look at star ratings, review volume, verified badges, and whether your business details match consistently across Google, Yelp, and the BBB. A contractor with a strong local reputation but a thin or inconsistent profile loses to a platform aggregating thousands of corroborated reviews.

Industry practitioners have noticed something counterintuitive: raw review count matters less than the depth of what each review actually says. One local SEO specialist observed that ChatGPT "keeps picking businesses with really specific, detailed reviews. Not more reviews. Not better ratings. Just... more words." They noted a client with 2,000 reviews and a 4.7-star rating never gets mentioned by ChatGPT, while a competitor with only 47 reviews gets cited every time because their reviews contain detailed paragraphs describing specific experiences.

The practical takeaway: your reputation has to be legible to a machine reading many sites at once, not just visible to a homeowner who already found you. A strong local name the wider web hasn't documented is, to the model, close to invisible.

How to Build Visibility Across AI Answer Engines

• Volume and Consistency: Ensure your business information matches across independent sources like Google, Yelp, the BBB, and local directories. Gaps and contradictions read as noise to AI systems and lower your odds of being recommended.
• Depth Over Brevity: Encourage detailed reviews that describe specific experiences, outcomes, and details rather than generic praise. AI models prioritize substantive review content when deciding which businesses to cite.
• Off-Site Documentation: Build presence beyond your own website through forum participation, third-party blog mentions, local press coverage, and industry listings. This expands your corpus share and makes you more discoverable to AI systems.

Why "Near Me" Doesn't Work in AI the Way It Works in Google Maps?

Most business owners assume AI local answers behave like the Google Maps pack, where proximity ranks results. They don't. AI resolves "best HVAC company in [city]" by disambiguating entities in the web corpus, matching a place name against what the web says about businesses there. There's no GPS proximity dial to turn, so you can't optimize for "near me" the way you do in the map pack.

There's a second surprise that cuts against the hype: AI answers appear far less often for commercial "hire someone" queries than marketing pitches suggest. In one research panel, Google AI Overviews appeared in only 16.5% of commercial queries with a location modifier and 19.9% without one, compared to 92% to 97% for informational queries. The queries that actually book jobs trigger AI answers less often than the pitch implies, a nuance almost no competing page discloses.

What Does an Angi Lead Actually Cost You?

While AI assistants recommend Angi without mentioning the cost, Angi's own 2024 SEC filing reports $587.1 million in US lead revenue, fees paid by service professionals for consumer matches. You're charged per match, whether or not you win the job. You're paying for a lead, not a customer.

HVAC leads run roughly $20 to $85 each, and the same lead is commonly sold to three to eight contractors at once. Shared leads close at 5% to 22%, while your own inbound calls close at 25% to 40%. One Texas HVAC owner shared a breakdown on Reddit that captures how the shared-lead model plays out: "Last summer I spent $3,400 on HomeAdvisor leads. Out of roughly 60 leads I closed 7 jobs. The rest either ghosted me, went with a cheaper competitor, or the number was disconnected. What really kills me is I found out those same leads were being sold to 4 other HVAC companies at the same time. So I am paying $80 for a lead, racing to call them first, and competing on price with 3 other guys who got the same lead. It is not a lead. It is an auction where they charge all of us to enter".

Before spending money on "AI optimization" services, measure your own market first. Anyone selling you certainty without testing your city first is skipping the only honest first step. The problem is real, but it's structural and lives mostly off your website, not on it.