Where Structured Data Standards Actually Come From
By the Numbers
Pew: AI summaries cut clicks from 15% to 8%.
Only 1% click a source cited inside AI answers.
Zero-click searches rose from 56% to 69%.
The advice about structured data is a mess right now. One study says it lifts your visibility in AI answers. Another says it barely moves the needle. This article cuts through that noise. It points you to the sources that actually set the rules, so your team can stop guessing and start building on something solid.
The problem
You have a small marketing team. Maybe two people. Maybe five. Nobody on it was hired to do SEO full time. And yet structured data keeps coming up in every meeting about AI search, because everyone has read that it might be the thing that keeps you visible.
So you go looking for answers. And the answers do not agree.
Vendors promise big lifts. The most rigorous study to date found something far quieter. Ahrefs tracked 1,885 pages that added schema between August 2025 and March 2026, matched them against 4,000 similar control pages, and measured the change. Citations barely moved on any AI platform. That is the kind of study you can actually trust: large sample, real control group, careful method. And its headline finding is not the one the loud advice is selling.
So you are left holding two stories at once. One says structured data is the key to staying visible. The other, and the better-built one, says adding it did almost nothing for pages that were already in the mix. And you are supposed to make a budget call based on that.
The stakes are not small either. The reason structured data is on your radar at all is that the ground under organic search is moving. Here is what that looks like in well-sourced numbers.
Pew Research tracked real browsing for 900 US adults in March 2025. When an AI summary appeared at the top of the results, people clicked a normal search result only 8% of the time. Without a summary, they clicked 15% of the time.
Only 1% of those visits ended in a click on a source cited inside the AI answer itself.
Similarweb data shows zero-click searches climbed from 56% to 69% between May 2024 and May 2025.
So the journey that used to end on your site now ends on Google's results page. The visitor gets the answer and never comes to you. Your instinct as a CMO is correct. This is a real shift, not a passing fad. The trouble is figuring out what to do about it when the people selling you the answer keep contradicting each other.
That is the real problem. It is not whether structured data matters. The problem is that you cannot tell whose numbers to believe, and you do not have a spare analyst to sort it out for you.
When every vendor has a different number, the only safe ground is the source that writes the rules.
Where the confusion comes from
Here is the honest version. The loud claims and the careful studies are measuring different things, on different sites, at different moments, using different definitions of what a "citation" even is.
Take the Ahrefs finding. Every page in that study already had strong AI citations before any schema was added. Those pages had almost no room left to climb, so adding schema did not push them higher. The researchers were clear about that limit. It does not prove schema is useless. It proves that for pages already in the mix, schema was not the lever that moved them. Most of the splashier claims you see do not come with that kind of honesty about their own limits. Many come from companies that sell schema tools. That does not make them wrong, but it does mean you should read them with one eyebrow raised. And the AI platforms themselves keep changing how they work, sometimes from one month to the next.
The platform makers are careful with their own words, which tells you something. In April 2025, Google's Search team said structured data gives an advantage in search results. In March 2025, Microsoft's Fabrice Canel said schema markup helps Microsoft's models understand content for Copilot. Both statements are useful. Neither one spells out exactly how the schema is used, and neither one promises the behavior will stay the same.
So the marketing version of structured data is loud and unsettled. The standard version is quiet and stable. You want the quiet one. That is the version you can build a process around without rewriting it every quarter.
The sources that set the standard
There are five primary sources worth trusting. Not blog posts about the sources. The sources themselves. Bookmark these. Treat them as the law, and treat everything else as commentary on the law.
1. Google Search Central, General Structured Data Guidelines
This is Google's official policy for all structured data used in Search. It defines what makes a page eligible for rich results, which formats are allowed, and where the lines are for spam and abuse.
To be eligible for rich results, you have to follow these guidelines and use a supported format. Google recommends JSON-LD.
Your markup has to match what a visitor actually sees on the page. Hidden, misleading, or incomplete schema can get a page disqualified.
That second rule is the one most teams break, and it matters more than people think. Several studies have found that schema which contradicts your visible content is worse than having no schema at all. If your markup says one price and your page shows another, the platform can throw out the whole thing. So the goal is not "more schema." The goal is accurate schema.
2. Google Search Central, Introduction and the Search Gallery
These two pages tell you which schema types Google actually uses, and how to put them in place. The Search Gallery is the official list of rich result types: Article, Product, FAQ, and the rest, each with instructions and validation guidance.
These pages are where Google says, plainly, to treat its own documentation as the final word on how Search behaves.
In Google's own framing, you should rely on its documentation as "definitive for Google Search behavior."
Why does that line matter so much? Because of the next source.
3. Schema.org, the official vocabulary
Schema.org is the shared dictionary. It was founded by Google, Microsoft, Yahoo, and Yandex, and it defines the types and properties that everyone else draws from.
It gives you machine-readable definitions of every type and property, plus release notes and downloadable files.
A formal community group runs the process of adding and changing schema over time.
Here is the catch, and it is worth saying out loud. Schema.org defines the words. Google decides how it reads them. When the two disagree, Google's documentation wins for Google Search. So you use Schema.org to learn the vocabulary, and you use Google Search Central to learn what Google actually does with that vocabulary. One is the dictionary. The other is the rulebook.
Schema.org defines the words. Google decides how it reads them.
4. Google Search Central, Preferred Sources and AI Features
These are the newer docs, and they are the ones a CMO worried about AI search should read first.
The Preferred Sources guide explains how a site can be highlighted in places like Top Stories, and what compliance it requires.
The AI features page explains that AI Overviews and AI Mode run on Google's normal indexing and eligibility rules. The same structured data and snippet rules that govern regular search.
Read that last point twice. Google is telling you that getting into AI answers is not a separate, secret game. It runs on the same plumbing as the search you already know. That is good news for a small team. It means you do not need to invent a second strategy from scratch. You need to get the first one right.
5. Google Merchant Center, structured data for products
If you sell anything, this is your standard for product schema.
Your structured data has to live in the HTML the server returns, and it has to match the price and availability shown on the page.
It spells out the required properties: price, currency, availability, and condition. JSON-LD is recommended.
This is the same "match what the user sees" rule from the first source, now applied to the thing that makes you money. Get it wrong and you do not just lose a rich result. You can lose your product listings altogether. For an ecommerce brand, that is not a missed opportunity. That is revenue walking out the door.
What this actually tells you
Step back and look at all five sources together. A pattern shows up fast.
Every one of them says the same quiet thing in a different way. Use a supported format. Match your visible content. Follow the documented rules. Do not try to trick the system. That is the whole standard, repeated five times by five different documents.
Now compare that to the loud advice. None of these official sources promise you a giant overnight lift. None of them sell you a shortcut. The official standard is boring on purpose. And boring is exactly what you want when the loud advice keeps changing under your feet.
The Ahrefs study makes the point in a way that is easy to misread, so let me be plain about it. For pages that were already being cited by AI, adding schema did not push them higher. That sounds like bad news for schema. But the same researchers noted that for pages not yet being seen at all, structured data may still help them get crawled, parsed, and understood in the first place. So here is the honest reading: structured data is table stakes, not a magic lever. It helps the system understand you. It does not bribe the system to like you.
There is one more thing worth sitting with. Most companies still have not done the basic version of this well. Read that as opportunity, not pressure. While your competitors chase the loud thing, you can quietly do the simple thing correctly. In a market full of noise, correct and boring is a real edge.
The takeaway
Takeaway 1: The structured data debate is loud because the marketing world is arguing about effect sizes. You do not have to join that argument. Anchor your team on the five primary sources, and the argument stops mattering to your budget.
Takeaway 2: Structured data is not a growth hack. It is the language that helps search and AI systems understand what your page is. Done right, it makes you eligible. Done wrong, it can disqualify you. There is no version where it is both optional and safe to ignore.
Takeaway 3: AI Overviews and AI Mode run on the same rules as regular search. You do not need a separate AI strategy. You need your existing structured data to be correct, current, and matched to what visitors actually see.
The action
Action 1: Make the primary sources your single source of truth. Bookmark the five sources in this article and tell your team to check them before acting on any blog post or vendor claim. When someone shares a stat about schema and AI, the first question is simple. Does Google's own documentation back this up? If not, treat it as commentary, not policy.
Action 2: Run your top pages through Google's Rich Results Test. Pick your ten most important pages. Run each one through Google's Rich Results Test and check the enhancement reports in Google Search Console. You are looking for two things: valid markup, and markup that matches what the page shows. Fix the mismatches first, because mismatched schema can hurt you more than missing schema does.
Action 3: Standardize on JSON-LD. If your site uses a mix of formats, move to JSON-LD. It is the format Google recommends across its documentation, and it keeps your markup separate from your page layout. That separation makes the whole thing far easier for a small team to maintain over time.
Action 4: Lock down product schema if you sell anything. Open Merchant Center's structured data requirements and confirm your product pages include price, currency, availability, and condition. Then confirm those values match the live page. This protects your listings, not just your rich results, so it pays for itself fast.
Action 5: Treat schema as the foundation, then move on. Get the markup right, then put your energy where it actually moves the needle. That means content that answers the real questions your buyers ask, and visible content worth citing in the first place. Structured data helps AI understand you. Good content is what makes you worth understanding.
The loud advice will change again next quarter. The standard will not. Build on the standard.