Nobody Wrote the Answer Down

There is a cell in the spreadsheet of every industry cohort we crawl, and it keeps coming back with the same value. The metric: of all the structured data a sector publishes (the JSON-LD and schema.org markup that describes pages to machines), what share is answer-shaped, the FAQ and question-and-answer formats an AI assistant can lift directly into a reply?

The value: approximately none.

The numbers, sector by sector

Our technology cohort set the high-water mark, and it was 6%: across ten Fortune 500 tech sites and roughly 21,000 schema annotations, 6.1% was answer-shaped, and most of that came from a single SaaS company whose distribution model relies on programmatic discovery and citation.

Then it got worse. The retail cohort produced zero: 5,073 annotations across eleven corporate sites, not one FAQ or Q&A entity among them. Industrial produced the largest schema corpus we have measured (23,823 annotations across fourteen sites), containing exactly 3 answer-shaped entries. Telecom and media managed 1.7%, almost all of it from one carrier’s support library, the lone case in two quarters of a company marking up content that actually answers customer questions.

What they wrote down instead

The near-zero is not a markup-adoption problem. Many of these sites mark up everything; we routinely measure 97-100% page coverage. The composition is the finding. The annotations describe organizations, addresses, breadcrumbs, page types, search boxes: a machine-readable org chart and site map. Schema, as deployed by the Fortune 500, is a business card.

That made sense for the reader it was written for. Search engines used identity markup for knowledge panels and breadcrumbs for result decoration. When traditional search engines curtailed FAQ visual snippets on result pages, many enterprise teams abandoned question markup entirely. That made sense when the consumer was a human scanning blue links. Answer engines invert the requirement: an assistant synthesizing a direct response wants structured claims (what does this plan include, what is the warranty, how does the process work, what does this company actually make). SERP decoration and DOM consumption are different jobs. LLM scrapers and RAG pipelines still read schema in the page for entity extraction even when rich results no longer show a FAQ accordion. The formats for stating those claims machine-readably have existed for years. Approximately nobody in three sectors is using them.

The mismatch is sharpest where the questions are

Rank the sectors by how many consumer questions AI assistants field about them daily, and the answer-shaped share runs almost exactly backwards. Retail (what’s the return policy, does this fit, which one should I buy) wrote down zero answers. Telecom (why is my bill higher, what does this plan include) wrote down 1.7%, and two thirds of the pages carrying it had no inbound links. The one sector with a measurable share, technology, got it mostly from the single company whose distribution model relies on programmatic discovery and citation.

That last case is the tell about causation. Answer-shaped markup appears where being machine-quotable is treated as a distribution requirement, and nowhere else. It is not hard, expensive, or exotic. It is simply nobody’s KPI.

What this does not say

Answer-shaped schema is not a golden ticket, and we are not claiming FAQ markup alone gets you cited. Engines synthesize from body text too; markup is a strong hint, not a contract, and its weight varies by engine and has shifted over time. Our cohorts are curated Fortune 500 samples, and our classifier counts schema types, not answer quality. The durable claim is comparative: the same companies that invested in near-total identity markup invested approximately nothing in the one schema family aimed at the questions assistants actually answer. Whatever the exchange rate of markup to citations, that allocation is wrong.

The cheapest gap in the series

Most findings in this series are structural and cost real work to fix: re-linking archives, rebuilding hierarchies. This one is closer to a writing task. The companies already employ people who answer these questions all day: support teams, product marketing, sales engineering. The answers exist as tickets, scripts, and internal wikis. What is missing is publishing them as pages that state one claim cleanly, and marking them up in the format built for exactly that.

And because approximately nobody in these sectors has done it, the first mover in each category gets a structural advantage: when an assistant evaluates candidate sources for an answer about your product class, machine-readable claims lower the extraction cost. You force the model to weigh clean first-party assertions against messy third-party consensus (Reddit threads, affiliate roundups, trade copy). We measure this gap, alongside whether anything links to the pages that would carry those answers (which is its own epidemic), in every Digital MRI. The machine-readable web your sector built says who everyone is. The one it forgot to build says what anyone can answer. That second web is, for the moment, almost empty. Being early in an empty field is rarely this cheap.