Why ChatGPT Recommends Your Competitor's HVAC Company Instead of Yours — Even Though You Outrank Them on Google
A homeowner in Redondo Beach types "best HVAC company for a mini-split install near me" into Google. Your site shows up at position #3, right under the map pack, with 140 reviews and a 4.9 average. Then Google's AI Overview loads above all of it — and it names a competitor with 38 reviews and a rank of #7.
This isn't a fluke, and it isn't Google being random. It's a mechanical, repeatable failure that's already happening to South Bay contractors, and almost none of them know it's happening because nobody checks. You check your map pack position. You check your organic rank. You don't check what ChatGPT, Perplexity, or Google's AI Overview says when someone asks it to just tell them who to call — because that interface doesn't show you your position. It just gives an answer, and either your business is in it or it isn't.
Ranking and Getting Cited Are Now Two Different Games
Traditional SEO optimizes for one outcome: rank high enough that a human clicks through, reads your page, and calls. That's a ranking problem, and it's mostly solved by backlinks, review volume, on-page keywords, and site speed.
AI search doesn't rank pages for a human to click. It extracts an answer and generates it directly. The model reads dozens of pages in the time it takes you to read this sentence, and it's not looking for the page that "deserves" to rank — it's looking for the page that hands it a clean, quotable fact it can lift without editing. If your service page says "We pride ourselves on quality craftsmanship and customer satisfaction, contact us today for a free estimate," there is nothing in that sentence an AI model can extract. It's persuasion copy, not information. The model skips it and moves to the next page — often your competitor's — because that page says "Mini-split installation in Redondo Beach typically runs $4,500–$8,500 depending on zone count and unit tier, and most jobs are completed in one day."
That second sentence is a fact. It has an entity (mini-split installation), a location (Redondo Beach), a value (price range), and a resolution (timeframe). It's citable. Your marketing copy isn't.
The Content That Wins Google Rankings Is Often the Content That Loses AI Citations
Here's the part that trips up owners who've already invested in SEO: the same page can rank well and get ignored by AI simultaneously, because Google's classic algorithm and its AI layer are scoring different things.
Classic ranking rewards topical depth, backlinks, dwell time, and review signals — all of which your 140-review, well-aged domain has in spades. AI extraction rewards structural clarity: does the page contain a directly answerable sentence sitting close to the question a user actually asked? Most contractor sites bury their answers three paragraphs deep, wrapped in brand voice, after a hero section about "decades of trusted service." The AI model has already generated its answer from a competitor's page by the time it would've reached yours.
This is the same failure mode showing up across home services businesses right now — sites built to persuade a human who's already on the page, with zero structure built for a machine that's deciding whether to mention the page at all.
Your FAQ Section Is Decoration, Not Infrastructure
Most trades sites have an FAQ block because a template came with one, or because a past SEO vendor said "FAQs help SEO." It's usually five soft questions like "Why choose us?" and "What areas do you serve?" with paragraph answers that hedge instead of state.
An FAQ section built for AI extraction does something different: it mirrors the actual phrasing people use when they ask a voice assistant or type into ChatGPT. Not "What services do you offer?" but "How much does it cost to replace a water heater in Torrance?" Not "Why choose us?" but "How long does a panel upgrade take for a house built before 1980?" Each answer needs to be a self-contained, extractable fact in the first sentence — location, service, number — with elaboration after, not before.
The Mechanism: AI Models Trust Structured Data More Than Prose
Underneath the writing problem is a technical one. Google's AI Overviews and most LLM-powered search tools lean heavily on structured data — Schema.org markup — to decide what a page is actually about and whether to trust it as a source. A Service schema block that explicitly defines areaServed, priceRange, and serviceType gives the model a machine-readable confirmation of what your prose is claiming. A FAQPage schema wrapping your Q&A content tells the model exactly where the extractable answers live, instead of forcing it to parse unstructured paragraphs and guess.
Most WordPress-built contractor sites either have no schema at all, or have generic LocalBusiness markup dropped in by a plugin with the wrong category, no areaServed field, and no service-level detail. The AI model isn't being unfair to these sites — it's making a confidence judgment, and a site with no structured confirmation of its claims loses that judgment to a site that has it, every time, regardless of who has more reviews.
What a Fixed Architecture Actually Looks Like
The businesses starting to show up in AI Overviews and ChatGPT answers right now share three structural traits, and none of them require a rebrand.
Every core service page states a direct, numeric answer near the top — price range, timeframe, or scope — before any brand language. "Trenchless sewer line repair in Torrance runs $6,000–$12,000 and avoids tearing up your driveway" beats "We're the trusted trenchless repair experts" on every AI extraction, every time, because the first sentence is a fact and the second is an opinion.
Every service-and-city combination gets its own FAQ block phrased in natural question form, wrapped in FAQPage schema, so the model can match a user's literal query to a literal answer on your site instead of stitching one together from a competitor's page and a Reddit thread.
Every page carries Service schema with explicit areaServed, priceRange, and serviceType fields that back up what the prose already says — so the model isn't trusting your claim on faith, it's confirming it against structured data twice.
This is exactly the layer most plumbing and trades websites never get built, because the agencies that built them were optimizing for a 2021 version of Google that no longer exists on its own — it now sits underneath an AI layer that reads differently than it ranks.
The Businesses That Get Cited First Are the Ones Still Getting Cited in Two Years
AI search doesn't work like organic rankings, where fourteen contractors can all sit on page one and split the traffic. It gives one answer, sometimes three. There's no page two. The South Bay HVAC company, electrician, or GC that gets extraction-ready first becomes the default answer for their service area — and every subsequent AI training pass reinforces that citation, because the model has already learned your business is the trustworthy source for that query. The ones who wait are optimizing for a search interface that's already splitting in half underneath them.
If you're running a home services company anywhere from Torrance down through the South Bay and you've never checked what ChatGPT or Google's AI Overview actually says about your trade in your city, that's worth twenty minutes before it's worth another dollar in ad spend. Ask it the question your customers are already asking. If your name isn't in the answer, the fix isn't more content — it's the right structure underneath the content you already have. That's a conversation worth having before your competitor has it first.