How to Write Service Pages That AI Models Actually Cite

Most service pages are written for humans who already decided to hire someone. The goal is to reassure them, list credentials, and prompt a call. That's fine for conversion — but it's not what gets you cited by an AI.

When ChatGPT or Perplexity needs to answer "who does kitchen cabinet painting in Oakville," it's looking for pages that directly, explicitly, and structurally answer that question. Most service pages don't. Here's how to build ones that do.

The fundamental difference in how AI reads your pages

A person scanning your service page will tolerate vague headlines and creative copy because context fills the gaps. An AI model has no patience for ambiguity — and in a competitive query, it will cite the business whose page makes its job easiest.

AI retrieval works by matching language in your page against the language of the user's query. If someone asks "licensed electrician for EV charger installation in Mississauga" and your page says "residential electrical services" with no mention of EV chargers or Mississauga, you won't match. Another electrician whose page says exactly those words will.

This isn't a keyword-stuffing exercise. It's about being explicit where you'd otherwise be vague.

Start with the page title and H1

Your H1 should name the service and the location. Not "Our Plumbing Services" — "Emergency Plumbing Services in Guelph, ON."

This feels blunt. It is. But an AI retrieving context for "emergency plumber Guelph" is going to have high confidence in a page that opens with exactly that phrasing. That confidence is what drives a citation.

If you serve multiple cities, you have two options: a single page that explicitly lists each city in the body text (weaker), or separate service-area pages for each city (stronger, and what we recommend for Perplexity in particular, which does live web search on every query).

Write the first paragraph for machines, not just people

The opening paragraph of your service page carries disproportionate weight for AI retrieval. It needs to state — plainly — what you do, where you do it, and for whom.

Compare these two openings for a landscaping company:

*"We believe a great outdoor space transforms how you live. Our team brings creativity and craftsmanship to every project."*

vs.

*"We provide residential landscaping services in Hamilton and Burlington, including lawn care, garden design, interlocking stone, and seasonal cleanup. We serve homeowners in Hamilton, Stoney Creek, Ancaster, Dundas, and Burlington."*

The second version is less poetic. It also tells every AI retrieval system exactly what query contexts this page matches.

Name your services the way people ask about them

The language you use internally for your services is often not the language a customer uses when asking an AI. The gap between those two vocabularies is where citations get lost.

A law firm might call it "Residential Real Estate Transactions." A first-time homebuyer asks ChatGPT: "real estate lawyer for buying a house in Kitchener." Your service page should include the phrase "buying a house" — not just the formal terminology.

Look at your actual service names and ask: how would someone with no industry knowledge ask for this? Write that phrase into the page, alongside whatever formal name you use.

Add Service schema to each individual page

Most businesses that have schema markup have it only on their homepage. That's better than nothing, but it misses a significant opportunity.

Individual service pages should have their own Service schema block — a JSON-LD script that explicitly names the service, links it to your business entity, and specifies the area where you offer it.

A minimal Service schema for a page looks like this:

```json { "@context": "https://schema.org", "@type": "Service", "name": "EV Charger Installation", "description": "Licensed EV charger installation for residential and commercial properties in Mississauga, Brampton, and Oakville.", "provider": { "@type": "ElectricalContractor", "name": "Your Company Name", "address": { "@type": "PostalAddress", "addressLocality": "Mississauga", "addressRegion": "ON" } }, "areaServed": ["Mississauga", "Brampton", "Oakville"], "serviceType": "EV Charger Installation" } ```

The `areaServed` array is where most implementations fall short. Without it, AI models don't know which cities this service covers — and they won't cite you for location-specific queries.

Use a FAQ section — and make it answer real AI queries

A FAQ section at the bottom of a service page is one of the best AI citation levers available. AI models are specifically trained to answer questions, so question-and-answer formatted content maps directly onto how they retrieve and generate responses.

The key is writing questions as people actually ask them, not as leading questions for a sales pitch.

Bad FAQ: *"Why choose us for your home renovation?"*

Good FAQ: *"How long does a bathroom renovation take in Hamilton?"* or *"Do I need a permit for a kitchen renovation in Ontario?"*

When Perplexity or Claude gets asked one of these questions, a page that has the exact question and a clear, specific answer has a structural advantage over a page that discusses the topic generally.

Aim for five to eight questions per service page. Keep answers under 100 words each — concise answers get cited more readily than paragraphs.

The single-page trap

Many service businesses have one page titled "Services" that lists everything they do. This is the format that performs worst for AI citations.

A single page trying to cover roofing, siding, eavestroughs, and window installation is optimized for none of those things. An AI retrieving context for "eavestrough replacement in Burlington" will not match that page confidently when a competitor has a dedicated eavestrough page with schema, a specific FAQ, and Burlington mentioned throughout.

Separate service pages also mean separate schema blocks, separate FAQs, and separate internal linking targets — each one a new entry point for AI retrieval.

If building out full individual pages isn't feasible right now, at minimum break your services page into clearly named anchor sections and add schema for each service type.

Internal linking connects the dots

AI crawlers follow links the same way search crawlers do. Your homepage should link explicitly to each service page. Each service page should link to related services.

A roofing company homepage that links to "Roof Replacement," "Roof Repair," "Flat Roofing," and "Emergency Tarping" is telling every crawler — including GPTBot and Perplexity's crawler — that these are distinct, substantive services. That structure contributes to how confidently an AI model describes your service offering.

If you have service-area pages, link to them from the relevant service pages. "We provide eavestrough replacement across Hamilton, Burlington, and Oakville — see our Burlington location page" is both useful for humans and meaningful for AI entity mapping.

What to fix first

If you have one service page that covers everything: split it. Pick your top two or three services and build dedicated pages first.

For each page: update the H1 to include service + city, rewrite the opening paragraph to be explicit, add Service schema with `areaServed`, and add a FAQ section with real questions.

These changes don't require a site rebuild. Most can be done page by page in whatever CMS you're using.

Not sure which of your service pages are showing up in AI queries — and which aren't? A Signal Check at sourcepull.ca runs 40 real queries across ChatGPT, Perplexity, Claude, and Gemini, and shows you exactly which services are getting cited versus ignored. It's the fastest way to know where to focus.