AI Search Mistakes That Keep Contractors Invisible to ChatGPT

Mistake 1: Assuming Google rankings carry over to AI answers

This is the mistake underneath every other one. You rank page one for "AC repair" in your city, the map pack has you in the top 3, and you assume the AI engines see the same thing and will name you the same way. They do not, and the gap is widening.

Classic search returns a list of links and lets the homeowner pick. Answer engines return a decision. ChatGPT, Gemini, Perplexity, Copilot, and Google's AI Overviews read the web, decide which businesses are relevant, and hand back a short named list with reasons. That is a different job with different inputs. A blue-link ranking is earned largely on your own pages and links. An AI mention is earned on whether the model can confidently identify your business, understand exactly what you do and where, and find other sources that back it up.

Here is the practical proof: run the query yourself. Open ChatGPT and ask, "Who are the best-rated HVAC companies in [your city]?" Then ask a follow-up, "Which ones do commercial rooftop units?" Watch what gets named. Owners who rank beautifully on Google routinely watch the model list three shops that outrank them nowhere, because those shops are easier for the model to identify and confirm.

There is a timing angle too. Homeowners who used to type "AC repair near me" now ask an assistant who to call, and they act on the short list it gives back. If a competitor is named and you are not, you never entered the consideration set. You did not lose the bid. You never got asked to bid. That is a quieter kind of loss than sliding down a rankings report, which is exactly why so many owners have no idea it is happening to them.

The takeaway is not that Google work is wasted. Your organic and local rankings still feed the picture, and the entity, schema, and corroboration work here strengthens them in return. The takeaway is that AI visibility is a separate layer with its own mechanics, and treating it as a free byproduct of SEO is exactly why established contractors are getting quietly written out of the answer that homeowners now read first.

Mistake 2: Your business is a fuzzy entity the model can't pin down

An answer engine has to answer one question before it can name you: is this a real, specific business, and what exactly is it? That is the entity problem, and it is where most contractors quietly fail.

To the model, a clean entity looks like this: one consistent legal name, one primary trade stated plainly, a defined service area, and a set of facts that never contradict each other across the web. A fuzzy entity looks like the opposite. Three different versions of your name (LLC on the invoice, DBA on the truck, a nickname on the reviews), a homepage that says "we do it all," a service area you never actually state in text, and a phone number that changes between your site, your directory listings, and your profiles.

When the facts conflict, the model does not guess. It hedges, and hedging means it reaches for a business it can describe with confidence instead. Common ways contractors blur their own entity:

• Using different company names on the website, invoices, review sites, and directory listings.
• Never stating the service area in plain words on the page (a map embed is not text the model reads as fact).
• Listing fifteen trades with equal weight so no single specialty is legible.
• A phone number and address that are inconsistent from one source to the next.

The fix is boring and it works: pick one name, one primary trade, one service area, and one NAP (name, address, phone), and make every public source say the identical thing. State those facts in plain text on your own site, not just in a logo or an image, because models weight text they can read. Then reconcile the rest of the web to match. Entity clarity is the foundation the rest of AI visibility is built on, and no amount of clever content saves a business the model cannot reliably identify.

One caution for owners with a broad shop. Being clear about a primary trade does not mean hiding the others. It means giving the model a spine to hang them on. "Electrical contractor serving [area]" that also does generators and EV chargers reads as one identifiable business with a specialty and add-ons. "We do everything" reads as noise. The model can name a specialist. It struggles to name a blur.

Mistake 3: Your pages read like brochures, not citable answers

Answer engines cite sources that directly answer questions. Most contractor sites answer nothing. They open with a hero line about quality and experience, a stock hero image, and a call to action, then bury the actual facts a homeowner (and a model) would want to know.

Think about what the model is doing. A homeowner asks, "How much does a metal roof cost in Tampa?" The engine wants a source that states a real answer in plain language it can quote. If your roofing page says "we deliver quality roofing solutions," there is nothing to cite. If it says the actual range, the factors that move it, and the local considerations, you become a citable source. That is the whole game.

Pages that get cited tend to share a shape:

• A plain-language answer to a real homeowner question, stated early on the page.
• Specific facts: ranges, timelines, materials, permit realities, what is and is not included.
• Clear structure the model can parse, headings that are actual questions, and short direct paragraphs.
• An at-a-glance block that lays out the essentials as a clean list a model can lift whole.

There is a fast way to find your gaps. Write down every question a homeowner asks you on a first phone call before they will book: what it costs, how long it takes, whether they need a permit, what happens if it rains, how you handle the old unit or the old roof. Now check whether your website answers each one in plain text. Most contractor sites answer almost none of them, which means the model has to source those answers from somebody else, and that somebody else gets named.

The mistake compounds when every page is a variation on "call us for a quote." The model has no fact to attribute to you, so it cites the competitor who published the number. You do not have to give away pricing you have not locked in. You do have to answer the questions homeowners actually ask, in text, on the page, in language a model can quote back. Brochures get skipped. Answers get cited.

Mistake 4: Missing or sloppy schema, so the model has to guess

Schema is the structured-data layer that hands facts to machines in a format they cannot misread. Blue-link SEO uses it for rich results (stars, FAQs, breadcrumbs). AI answer engines use it for something more basic and more important: parsing who you are and confirming what your page claims. This is the schema that matters most for being named, and most contractors either skip it or ship it broken.

The point is not decoration. When your page says in prose that you are an HVAC contractor in Naples, and your schema independently states the same business name, the same trade, the same service area, and the same phone in machine-readable form, the model gets corroboration from a single page. Facts that agree in two formats read as trustworthy. The kinds of schema that carry weight for AI parsing:

The sloppy version is worse than none. Schema that names a different business than the visible page, a stale phone number, a service area that contradicts your text, or FAQ markup that does not match the on-page answers byte for byte all read as untrustworthy, and untrustworthy sources get dropped. This is a common own-goal: a template plugin fired years ago, the business changed, and now the machine-readable facts quietly disagree with the visible ones. The model notices even when you do not.

Done right, schema is the cheapest, most reliable way to remove the model's uncertainty about who you are, and it is a one-time build that keeps paying off as long as you keep it honest. Note that schema for rich results and local ranking lives in your broader SEO and Local SEO work. Schema built specifically to be parsed and cited by AI, matched exactly to citable on-page answers, is the piece this layer owns.

Mistake 5: Nothing outside your own site backs up your claims

You can say you are the best plumber in town all day on your own homepage. The model discounts it. Answer engines weight independent corroboration far more heavily than self-description, because a business talking about itself is the least reliable source there is. This is where a lot of otherwise-solid contractors lose the AI answer: the whole story lives on their own domain and nowhere else.

Corroboration is the web agreeing with you when you are not in the room. It is your name, trade, and service area appearing consistently across the sources models actually trust, and it is other people describing what you do. When a homeowner asks the model for a recommendation, the shops that get named are the ones the model can confirm from more than one direction.

What corroboration looks like in practice:

• Consistent listings on the directories and platforms models read, with identical NAP everywhere.
• Reviews that describe specific work in specific places ("replaced our sewer line in [neighborhood]") rather than generic praise.
• Mentions on local and industry sources that are not your own site.
• Verified profiles that all point back to the same business, so the model can connect the dots.

Reviews deserve their own note here, because most contractors ask for them the least useful way. A five-star rating with no words tells the model nothing citable. A review that names the actual job, the actual neighborhood, and the actual trade gives the model specific, verifiable language to attribute to you. You cannot script that, and you should never try, but you can ask the customer who just watched you replace a panel in a specific part of town to say what you did and where. Specific truth corroborates. Generic praise evaporates.

The core mistake is treating your website as the entire proof. A model that finds you described only by you, in one place, has nothing to cross-check, so it leans toward a competitor the wider web confirms. You do not fabricate any of this, and you never should. You make sure the true story of your business is consistent and legible everywhere it already lives, and you give the web accurate, specific things to say about you. Corroboration is earned, not bought, and it is the single biggest lever on whether an AI names you.

Mistake 6: No way to know if you're even being mentioned

The last mistake is flying blind. Contractors will spend on their site and their rankings for years and never once check whether an AI engine names them, because there is no dashboard that tells them. Google Search Console does not report ChatGPT mentions. Your rank tracker does not watch Perplexity. So the channel that increasingly decides which contractor a homeowner calls runs completely unmeasured.

You cannot fix what you cannot see, and the reverse is true too: you cannot prove progress on AI visibility without checking the actual answers. The good news is that the free version of checking is available to you right now.

• Pick the five or six questions a homeowner would actually ask before hiring your trade in your city.
• Run each one through ChatGPT, Gemini, and Perplexity, and write down who gets named and what the model says about them.
• Note whether you appear at all, where, and whether the description is accurate.
• Re-run the same questions on a schedule, because model answers move.

Two things surface fast. First, whether you are in the answer or invisible. Second, and this one stings, whether the model is describing you wrong: naming the incorrect service area, an old phone number, or a trade you gave up years ago. A wrong mention can cost you more than no mention, because it sends a ready-to-buy homeowner somewhere you do not serve or to a line that no longer rings. When you find a wrong description, it usually traces straight back to one of the earlier mistakes: a fuzzy entity, a stale schema field, or a source on the web still carrying old facts.

Tracking is not a nice-to-have at the end of the list. It is how you find out which of the first five mistakes is actually hurting you, and it is how you prove the fixes worked. Do it by hand for now if you have to, on the same schedule you would check any other number that decides whether the phone rings. Without it, AI visibility is guesswork. With it, it becomes a channel you can actually manage, defend, and grow before your competitors realize it exists.