Schema markup for service businesses, in plain English.

Told versus inferred

Every machine that reads your website — search crawler, AI engine, answer box — is trying to work out facts. Who is this? What do they sell? What does it cost? Who is it for?

Without schema, the machine infers those facts from prose. It reads your beautifully written paragraphs and guesses. With schema, the machine is told. Schema markup is a small block of code — the standard format is called JSON-LD — that sits in the page and states the facts in a shared vocabulary: this is a business named X, it offers a service named Y, the price is Z, these are the questions people ask and here are the answers.

Inference fails quietly and often. A machine that has to guess your price says nothing, or worse, says a number from a page that isn't yours. Schema replaces the machine's guess with your statement.

The five types that matter for expert services

Schema.org defines hundreds of types. A service business needs about five. Past these, you're usually decorating:

• Organization or Person. The anchor: who runs this site, what it's called, where else it exists (LinkedIn, other properties). Every other type points back to this one. For a founder-led business, Person is often the honest choice — buyers are hiring you.
• Service. One per offer, on that offer's own page: what it is, who it serves, what it produces. This only works cleanly if each offer has a page — which is its own fix, covered in one page per service.
• Offer. The price, attached to the Service. This is the one founders resist and the one machines value most: a stated price is a citable fact. If your pricing is public in prose, putting it in schema costs you nothing and makes it machine-quotable.
• FAQPage. The questions buyers actually ask, with your actual answers. FAQ schema is how your answers — not a competitor's, not a guess — get pulled into answer boxes and AI responses.
• Article + BreadcrumbList. On your guides and long-form pages: who wrote it, what it covers, where it sits in the site. This is what makes a content layer citable rather than just readable.

Three myths worth clearing

1. "Schema is a ranking hack." It isn't, and treating it like one leads to stuffing markup with claims the page doesn't make. Engines check. Schema that contradicts the visible page gets ignored, and sustained mismatch erodes trust in everything else you mark up. The rule is simple: schema states what the page already says, in machine vocabulary.
2. "My platform handles it." Partially, at best. Most platforms emit generic WebPage markup — the machine equivalent of a shrug. The types that carry business facts (Service, Offer, FAQPage) almost always require deliberate setup, because only you know the facts.
3. "It's too technical to bother with." Checking it is not technical at all — two free tools, five minutes, below. Writing it is a small, well-defined development task. It is nothing like the cost of a redesign, and it usually outperforms one — a point worth reading before any redesign.

The five-minute check

1. Paste your homepage into Google's Rich Results Test. Free, no login. It lists every schema type it finds. "No items detected" means machines are inferring everything about your business.
2. Paste your main service page into validator.schema.org. This one validates against the full vocabulary, not just what Google surfaces. Look for whether your offer and price appear as data, not just prose.
3. Read what's there against what's true. Stale schema is worse than none — an old price or a dead offer stated confidently in machine vocabulary will be repeated confidently by machines. Schema is part of the site that has to move when the business moves, which is a maintenance rhythm, not a one-time task.

Schema is one signal of several that decide whether machines can work with your site at all — the full picture, including rendering and crawl access, is in can AI read your website.

What it looks like on a real page

Take a diagnostic offer — call it an operations assessment, $2,500, three weeks, for founder-led firms. The page already says all of that in prose. The schema for that page states the same facts as data: a Service named "Operations Assessment," provided by your Person or Organization, with an Offer whose price is 2500 and whose currency is USD, an audience description, and a FAQPage carrying the four questions every prospect asks about it.

Nothing new was written. No claim exists in the markup that a visitor can't see on the page. That's the whole discipline — and it's why the work is small. A developer with the facts in hand marks up a typical five-offer site in a day. The hard part was never the code; it was having offers defined clearly enough to state as facts. If your offers resist being stated that plainly, the site is reporting an operating problem, not a markup problem.