Unpopular Opinion: Schema.org Is How Search Companies Offload Their Technical Debt onto the Rest of the Web

Wed, 17 Jun 2026 00:00:00 +0000

A Provocation, Not a Manual

Let me say the quiet part out loud: for most websites, in 2026, Schema.org structured data is a transfer of cost from trillion-dollar search and AI companies onto you.

It’s marketed as a best practice. It’s sold as “helping the machines understand your content.” And in a handful of cases, it genuinely is. But strip away the SEO folklore and look at the incentives, and you’ll see something less flattering: you do unpaid engineering work so that the most sophisticated parsing infrastructure ever built doesn’t have to.

This is a deliberately sharp take. Take it as a provocation worth arguing with, not a checklist to copy.

The FAQ Trigger

This whole argument started, as the best ones do, in a chat. A colleague suggested adding FAQ structured data to new FAQ pages “regardless of Google’s announcements.” Reasonable. Common advice. It takes little time, so why not make the machines’ job easier?

Here’s my answer: because “why not” is exactly how technical debt accumulates.

And the timing makes the point for me. Back on May 8, Google announced it was removing the FAQ rich result for almost everyone — it now surfaces only for a narrow set of well-known, authoritative government and health sites (see Google’s own changelog). Read that carefully: countless sites added FAQ structured data precisely to earn that rich result, maintained it, refactored around it — and then, with a single changelog entry, the payoff evaporated. The markup stays in your codebase. The benefit left the building. That’s the asymmetry in one clean example: you carry the cost forever, the platform withdraws the reward whenever it suits them.

If your questions and answers are written to help a human being, they already work. They render. They read well. They’re useful. The moment you wrap them in a second, parallel, machine-only representation of the same content, you’ve created a maintenance liability that pays you back in nothing and pays Google back in cheaper extraction.

Who Actually Benefits

Let’s be precise about who gains what.

When you mark up an FAQ for rich results, the most likely outcome is that the answer is lifted, displayed in the SERP or in an AI summary, and the user never visits your page. That’s a zero-click outcome, and increasingly a negative-click one — the search the user never even had to make.

So follow the value:

The search engine gets clean, pre-structured data it would otherwise have to extract. Cheaper compute, cleaner answers, more confident summaries.
The AI agent gets a tidy machine-readable payload to recombine into its own response, with no obligation to send anyone back to you.
You get… the privilege of having been parsed.

These companies employ brilliant engineers and run the most advanced natural-language parsing systems on Earth. Let them work. Write your content for people. If the machines want to read it, decode it, and resell it, they can spend their own compute doing so. We’ve already given them plenty.

The Fragility Nobody Prices In

The cost argument is the philosophical half. The fragility argument is the practical one, and it’s the part that actually burns companies.

Structured data is exposed client-side, even when it’s generated server-side. That means it lives outside the default ecosystem of the web. It is not the content. It is a shadow of the content, sitting in the <head> or in a JSON-LD block that no human ever sees and that no ordinary editorial workflow protects.

Now picture the realistic failure mode:

A team refactors the site’s content.
The person doing it is a competent web professional — but not steeped in technical-marketing minutiae.
They rewrite the visible copy and, with no warning, the structured data silently goes stale, contradicts the page, or vanishes entirely.

No error. No test failure. No alert. The page still looks perfect. The markup quietly lies to crawlers for months. And the bill to find and fix it all lands on the business, not on Google.

It gets worse at the architectural level. Because this data isn’t a native part of the web’s content layer, it’s the first casualty of any major rebuild. The brilliant friend-of-the-manager who won an imaginary Nobel Prize for dynamic React apps proposes turning the site into a single-page web app — and the whole structured-data apparatus gets torn out along with everything else. You were never holding an asset. You were holding a maintenance subscription with no off-switch.

So When Should You Use It?

I’m not a structured-data nihilist. There’s a clear, defensible boundary, and it follows one rule: use Schema.org for stable public facts that don’t cost you commercial visibility.

Specifically:

Organization and contact data — the org chart, contact channels, opening hours, physical location. Information that rarely changes and that you want surfaced instantly.
Genuinely procedural content — recipes are the canonical example: structured, finite, and where the markup maps cleanly to the content.
Public administration — almost everywhere. When the goal is to serve citizens with public, non-commercial information as efficiently as possible, structured data is a public good. Mark it all up.

For private businesses, keep it narrow: use it to speed up retrieval of your contact and company information, and stop there.

And the litmus test for everything else is brutally simple:

Every zero-click that strips away commercial visibility is a tumor to be nuked from orbit.

If structured data adds visibility — great, use it. If it leaves visibility unchanged — fine, your call. If it removes visibility while shifting cost and risk onto you — don’t.

The Bottom Line

Schema.org isn’t evil. It’s a tool, and like any tool its value depends entirely on who’s paying the cost and who’s collecting the benefit. For public, stable, non-competitive facts, it’s a fair trade. For your hard-won commercial content, it’s increasingly a deal where you supply the labor, you carry the fragility, and someone else keeps the customer.

Write for people first. Make the search and AI companies earn the rest. They can afford it.

This post grew out of a Bluesky thread of mine — come argue with me there.

Sources

Google Search Central. “Removing the FAQ rich result (May 8).” developers.google.com — Search status and changelog.

Schema.org | Andrea Scarpetta >> 2025++