A product question has a right answer. A service question has a suitable answer, and that difference changes what the models look for.
The shape of a service query
Nobody asks an assistant for "a plumber". They ask about a leaking hot water system on a Sunday, in a neighborhood, with a budget worry attached.
Three things are embedded in that: a specific job, a location, and a constraint. The model is matching a business to all three, which is a different operation from ranking pages about plumbing.
The consequence is that specificity beats authority more often here than in product categories. A small business that has clearly stated which jobs it does and where wins matches that a larger, vaguer competitor does not.
Which sources decide it
For most service categories, the models lean on:
Review aggregates, heavily. With little brand recognition to work from, review volume and recency are the strongest available proxies for whether a business is real, active and reliable.
Category-specific registers. Licensing bodies, trade associations, insurer directories. These are treated as verified rather than claimed, and they are the ones most businesses have not completed.
Local editorial and directories, for anything with a geographic component.
Your own website is rarely the primary source. It earns its place by carrying the specifics those sources omit.
What to publish
The job, not the trade. A page about "emergency hot water repairs" answers a real question. A page about "our plumbing services" answers nobody's.
Your service area as text. A model cannot read a coverage map. It can read a sentence listing the neighborhoods you serve, and that sentence is what puts you in a local answer.
What things cost, at least indicatively. Price opacity is the defining anxiety in most service categories, and it is the question people most want answered. A range with the assumptions stated is unusual enough to become the passage a model quotes.
Who you are not for. Models reward sources that help them exclude. Saying plainly which jobs you do not take gets you quoted more and improves the fit of the people who do arrive.
What to measure
The same two columns as anywhere: were you named, and were you cited. But weight the prompt set differently.
Service queries are dominated by the buying moment rather than the research phase, so a tracking set for a service business should be almost entirely job-and-location questions phrased the way someone would say them out loud, with one or two brand-name prompts to see what is being said about you.
Ten prompts is enough. Track them fortnightly, logged out.
The advantage service businesses have
Product categories are contested by brands with budgets. Service categories are contested by local competitors who mostly have not done any of this.
Claiming your category-specific registers, writing five pages that answer real jobs, and stating your service area in words is a few weeks of unglamorous work, and in most service categories it puts you ahead of nearly everyone. That is not true in B2B software, where everyone has already done it.
The disadvantage
Reviews matter more, and they are the hardest input to manufacture honestly. A business with a thin or aged review profile reads as a risk to a model matching someone to a job, and there is no content strategy that substitutes for it.
The only real answer is the slow one: ask, consistently, and make it easy.
The service-business specifics
A service business has one structural advantage in AI search and one structural problem, and both are worth naming.
The advantage: your buyers ask questions, not keywords. Service purchases start with a described problem rather than a product name. "My hot water system is leaking and I rent" is exactly the shape of input these systems handle best, and it is exactly the shape of thing your team answers on the phone every day. That raw material is genuinely valuable and almost nobody writes it down.
The problem: you have very little indexable evidence. A product business has listings, specifications, reviews and comparison pages. A service business often has a five-page website and a phone number. Engines build answers from evidence, and there is not much to build from.
The work follows directly from those two facts. Write down the answers you already give verbally, one question per page. Then get the evidence that does exist, your registrations, accreditations, review profiles and completed work, into the sources engines read.
The order that works: claim and complete the institutional sources first, because they establish that you are real. Then the question pages, because they establish what you are for. Then third-party presence, because it establishes that someone other than you says so.
What it produced for a service business
“We've always grown through referrals - builders who know us pass our name on. That works, but it only reaches people who already know someone in the industry. Within a couple of weeks of the content going live, we had someone contact us directly through the website. That's a channel we didn't have before - and the enquiries have kept coming.
PROCERT provides building certification in New South Wales and had grown entirely on referrals. Referrals work, but only reach people who already know someone in the industry. Within a couple of weeks of the work going live, a stranger inquired directly through the website, and the inquiries kept coming.
Published with the starting number, the end number and the period: PROCERT.
A worked example
A commercial plumbing company, eighteen staff, half maintenance contracts and half reactive callouts.
The raw material problem. Their website was five pages: home, about, services, gallery, contact. Everything a buyer would want to know existed only in the heads of two estimators who answered the same questions on the phone forty times a week.
What they did first. Recorded those phone answers. Not wrote, recorded, then transcribed. "How fast can you get someone out for a burst main", "what does a backflow test cost", "do you work after hours", "can you do a compliance certificate the same day".
Why that worked. The transcripts were already in the customer's vocabulary, which is the shape a retrieval system matches against. Written from scratch they would have come out in trade language.
The evidence half. License register entry corrected. Two trade review platforms claimed. Association listing completed. About eight hours across three weeks.
Where it landed. Named on three of ten tracked questions by month two, all three being the specific job questions rather than the broad "best commercial plumber" one.
The general lesson. A service business is rarely short of expertise. It is short of that expertise existing anywhere a machine can read.
See which questions name you
A free AI visibility audit on your own buying questions, across all five answer engines.
Book your audit
Ashur Homa
Built and scaled a digital brand to $100M+ in sales with zero ad spend. Has helped businesses generate millions through AI go-to-market strategy. Leads growth at Omni Eclipse.
Connect on LinkedIn
