What Schema Implementation Costs — and What It Buys

Schema implementation costs whatever you pay to produce the JSON-LD once, multiplied by how many times you have to produce it. Generate it from data you already keep per page and the cost is one engineering project with a marginal cost near zero. Hand-author it, or buy it per page, and it becomes a variable cost that scales with your library forever and never stops.

That is the whole model. The harder question — whether schema buys anything at all in AI search — is settled evidence, and it is settled against the sales pitch: the verdict page is does schema markup help AI citations. This page prices what that page decides.

What are the actual line items?

Six, and only two of them recur.

The row that decides everything is per-page authoring. If markup is generated from fields that already exist — title, question, answer, dates, author, citations — then adding a page adds no schema work at all. Across our three builds, one frontmatter block per page emits FAQPage, Article, DefinedTerm and HowTo markup across 400+ published pages, and page 400 costs exactly what page 4 cost [our data].

What does the generated approach cost in practice?

A one-time engineering project, then nothing. The illustrative hours below are a worked example rather than a measurement we took; the specific hours and your wage assumptions are yours to replace.

Illustrative one-time build: type selection and mapping 3 hours, emitter implementation 4 hours, validation gate 2 hours, verification pass across existing pages 3 hours. Total: 12 hours. At a fully loaded $50 an hour that is $600; at $100 an hour, $1,200. Marginal cost per page after that: 0 hours.

Now the comparison that matters. At an illustrative 0.2 hours of per-page markup authoring, a 200-page library costs 40 hours — more than three times the one-time build — and a 400-page library costs 80. The crossover in this illustration arrives at 60 pages. Any library planning to pass that count is buying the wrong product when it buys markup by the page.

What does schema actually buy, given the evidence?

Machine-readable facts and eligibility for search appearances that require markup. Google's structured-data documentation describes exactly that transaction: markup tells Google what a page's content means, and certain appearances depend on it.

What it does not buy is AI citations. Ahrefs analyzed 1,885 pages between August 2025 and March 2026 and found no causal relationship between schema markup and AI citations — the strongest controlled test the niche has, and it points the other way from every pitch that sells markup as a citation lever. Price schema accordingly: as infrastructure with a fixed cost and a real but bounded payoff, not as a growth investment.

That bound is also why per-page schema pricing is hard to justify. A recurring fee attached to an outcome the evidence does not support is a recurring fee attached to nothing in particular.

What is the cost nobody quotes?

Retrofit. Markup decisions are library-wide, so changing one means touching every page that carries it. On our insurance build an early markup choice — a Q&A type applied to editorial question pages — was wrong for the page type and had to be reverted library-wide to Article plus FAQPage, with the primary question folded into the page instead [our data]. We never tracked the hours on either pass, so we cannot price the difference — only note the structural one: the original implementation ran once against an empty site, and the correction had to touch every page already published.

Validation is the second unquoted cost, and it is cheap only when automated. The day-one adversarial review on our auto-finance rebuild logged JSON-LD escaping as a defect to fix, alongside the output gate built to catch that class [our data] — a defect that source-level checks had passed. Budget the gate once and the checking is free thereafter; budget nothing and you are paying a person to eyeball markup on every release.

Both costs argue for the same thing: make markup a property of the system, not of the page. Where that leads for definitional pages specifically is DefinedTerm schema strategy.

When is spending on schema the wrong call?

Three cases. A site under about 20 pages, where the one-time build outweighs the benefit and hand-writing two JSON-LD blocks is genuinely faster. A site whose pages are not crawlable in raw HTML — fix that first, because markup on a page a crawler cannot read is markup nobody reads. And any site being sold schema as an AI-visibility service, where the correct response is to ask what evidence the seller has, given that the published evidence shows no causal link.

Said against our own interest: schema is the cheapest thing we build and the easiest thing to over-charge for. If a quote separates "schema markup" into its own recurring line, ask whether you are buying a template you keep or a fee that repeats. The per-page labor model that same question applies to is what a fact-audited page costs, and the wider argument for treating markup as infrastructure is in our generative engine optimization guide.

If you are trying to work out whether the whole library — markup included — is worth building in your niche at all, the fit check is the cheap way to find out.