Which question type matters most?

Problem-led questions. Someone typing "my AC is blowing warm air, who should I call in Phoenix" has an active problem, an open wallet, and no shortlist. Generic "best of" questions come from people earlier in the process who may be researching for later.

Why test branded questions if the customer already knows my name?

Because that is the reference check. A homeowner who got your name from a neighbor will ask an assistant whether you are legitimate. If the AI says it has no information about you, or gets your details wrong, you can lose a job you had already won.

Should I write content for all 25?

Write for the problem-led ones first — they map directly onto pages you can own, and they are the least competitive. Generic and comparative questions are won mostly through directory presence and reviews rather than through your own pages.

How often should I test these?

Monthly is plenty for most businesses. AI answers drift as sources change, and monthly testing catches a slide before it costs you a season.

Guides/Foundationsupdated 2026-08-04 · 4 min

The 25 Questions Customers Ask AI Before Hiring a Contractor

The short answer

Homeowners ask AI assistants four kinds of questions before hiring: generic local ("best plumber in Tampa"), problem-led ("my water heater is leaking, who do I call"), branded ("is Carter Plumbing any good"), and comparative ("top 5 roofers in Mesa, ranked"). Most businesses are invisible on the first two, which are the ones with buying intent.

Before a homeowner calls anyone, a growing share of them ask an assistant first. The questions are not random — they fall into four clear patterns, and each one tests something different about your visibility.

These are the twenty-five we run in our scan, and what each category tells you.

Category 1 — Generic local (the "best of" question)

The classic. Someone knows they need a service and wants a shortlist.

Who are the best [trade]s in [city]?
Best [trade] near me in [city]
Top rated [trade] in [city], [state]
Recommend a good [trade] in [city]
Which [trade] should I hire in [city]?

What it tests: whether you exist at all in the AI's picture of your local market.

What wins it: directory presence, review volume and recency, and appearing on the roundup pages the assistant reads. This category is won off-site, not on your website. If you are missing here, the fix is Yelp, Angi, BBB, Google, and getting onto the local "best of" lists.

Reality check: this is where most businesses score zero, and it is the category owners assume they are winning.

Category 2 — Problem-led (the highest-intent question)

Someone has an active problem and is describing it in their own words.

My water heater is leaking, who do I call in [city]?
I have a burst pipe in [city] — who can come out today?
My AC is blowing warm air, who fixes that in [city]?
There's a leak in my roof after the storm, who should I call in [city]?
My breaker keeps tripping, who do I call in [city]?
Emergency [trade] in [city] open now
Who does same-day [service] in [city]?

What it tests: whether the AI associates you with a specific problem, not just a category.

What wins it: content on your site that names the problem in the customer's words and answers it directly, plus reviews that mention that specific job, plus explicit availability language ("24/7," "same-day," "emergency").

Why it matters most: these people are buying today. And they are the least competitive, because most contractors write "Water Heater Services" instead of "My water heater is leaking — what do I do?"

Category 3 — Branded (the reference check)

Someone already has your name — from a neighbor, a truck, a yard sign — and is verifying you.

Is [business name] in [city] any good?
Tell me about [business name]
[business name] reviews
Is [business name] licensed and insured?
How long has [business name] been in business?
[business name] vs [competitor name]

What it tests: whether the AI knows you exist and can describe you accurately.

What wins it: a consistent public record. Matching name, address, and phone everywhere; license and insurance stated in plain text; years in business stated; substantive reviews.

The quiet risk: a homeowner who already chose you asks this question and hears "I don't have information about that business." That is a lost job you never knew you had.

Worth testing on yourself right now. Many owners have never done it.

Category 4 — Comparative (the shortlist question)

Someone wants a ranked list or a head-to-head.

Top 5 [trade]s in [city], ranked
Compare the best [trade]s in [city]
Who is the most affordable [trade] in [city]?
Which [trade] in [city] has the best reviews?
Best [trade] for [specific service] in [city]
Most reliable [trade] in [city] for emergency work
[trade] in [city] with the fastest response time

What it tests: whether the AI has a reason to rank you — a distinguishing attribute it can point at.

What wins it: having something specific to be best at. "Most affordable" needs price transparency somewhere public. "Fastest response" needs response-time claims and reviews confirming them. "Best reviews" needs the reviews.

A business with no stated specialty cannot win a comparative question, because there is no axis on which to place it.

Reading your own results

Run some version of these for your business and trade. The pattern of failure tells you the fix:

Failing category	The likely cause	Where to work
Generic local	You are not in the sources the AI reads	Directories, reviews, local roundups
Problem-led	No content in customer language	Your website, review wording
Branded	Inconsistent or thin public record	NAP consistency, credentials, About page
Comparative	No distinguishing attribute	Positioning, specialty pages, pricing transparency

Most businesses fail generic and problem-led while passing branded, which produces the specific illusion that things are fine — the AI knows who you are when asked directly, so it feels like you are visible. You are not. Nobody asks about a business they have not heard of.

What to do this week

Pick your five most common phone-call problems. Write each as a heading on your site with a forty-word direct answer.
Ask an assistant the branded questions about yourself and note every error.
Choose one thing you are genuinely best at and state it publicly in one sentence.
Ask five recent customers to name the specific job in their review.

Running all twenty-five

Doing this by hand across five assistants means a hundred and twenty-five prompts and a spreadsheet. Our free scan runs all twenty-five questions across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews at once, grades each assistant separately, and shows you every verbatim answer plus each competitor named in your place.

About fifteen seconds, free, no account required.

Common questions

Which question type matters most?: Problem-led questions. Someone typing "my AC is blowing warm air, who should I call in Phoenix" has an active problem, an open wallet, and no shortlist. Generic "best of" questions come from people earlier in the process who may be researching for later.
Why test branded questions if the customer already knows my name?: Because that is the reference check. A homeowner who got your name from a neighbor will ask an assistant whether you are legitimate. If the AI says it has no information about you, or gets your details wrong, you can lose a job you had already won.
Should I write content for all 25?: Write for the problem-led ones first — they map directly onto pages you can own, and they are the least competitive. Generic and comparative questions are won mostly through directory presence and reviews rather than through your own pages.
How often should I test these?: Monthly is plenty for most businesses. AI answers drift as sources change, and monthly testing catches a slide before it costs you a season.

Stop guessing

See exactly what the five AIs say about your business.

Twenty-five real customer questions across Claude, ChatGPT, Gemini, Perplexity, and Google AI Overviews. Free, about fifteen seconds, no account.

Run the free scan