Does Structured Data Still Matter Now That AI Writes the Answers?

It is a fair question, and I get asked it more than almost any other one right now. The rich results that schema used to earn you are disappearing. AI writes the answer at the top of the page before anyone scrolls. And Google itself has said, over and over, that structured data was never a ranking factor. So if it does not lift your rank and it no longer buys you a fancy snippet, why keep bothering?

Here is my actual take. Structured data matters more now than it did five years ago, not less. The reason it matters has flipped, though, and most people are still defending it for the old one.

The skeptics are not wrong about the snippet

Let me give the doubters their due, because the surface evidence is on their side.

First, schema was never a ranking signal, and that is not my opinion. Google’s own guidelines spell out that a structured data problem can cost you rich-result eligibility but does not affect how the page ranks in regular search. Adding markup has never been a lever you pull to climb the results. Anyone who sold it that way oversold it.

Second, the visible payoff really is shrinking. Google has spent the last couple of years retiring rich result types. HowTo results are gone. FAQ results have been wound down to almost nothing. A 2025 cleanup pulled several more. If your whole reason for marking up a page was a star rating or an expandable FAQ in the listing, that reason is genuinely fading, and no amount of insisting otherwise brings it back.

So if structured data were only ever about decorating your search result, I would tell you to deprioritize it and move on. It was never only about that.

Schema was never really about the snippet

The rich result was the visible byproduct. The actual job of structured data is translation.

Your web page is written for a human. It is full of implication, tone, and context a person fills in without noticing. A machine fills in none of that. Structured data takes the fuzzy human version and restates it as flat, unambiguous facts: this is a plumber, in this city, open these hours, offering these services, run by this person. Google’s own documentation is clear that it uses structured data not just for snippets but to understand the page and gather information about the companies described on it.

That job, making your business legible to a machine, did not get less important when AI started writing the answers. It became the entire point.

Why it matters more in an answer-engine world

Think about what an answer engine is actually doing when it decides whether to name you. It is trying to be confident enough about who you are to put you in a sentence and stand behind it. I wrote a whole separate piece on why that confidence is the real currency now. Structured data is the most direct way there is to hand a machine that confidence.

Unstructured prose makes the machine guess what you are. Structured data tells it, in the one format built for the job. JSON-LD is clean, predictable, and trivially parseable, which is exactly what you want when the reader is a model and not a person. It also feeds Google’s own entity understanding, the knowledge layer that sits underneath AI Overviews and Gemini, so the same markup that does nothing for your blue-link ranking can genuinely shape how AI systems describe you. Google has said as much directly, calling structured data critical for modern search features and the AI results that depend on clear, machine-readable content.

Notice what happened there. The thing that is dying, the decorative snippet, and the thing that is growing, machine comprehension, were always two different jobs the same markup quietly did. One of those jobs is shrinking. The other just became the main event.

What this should change about how you use it

The takeaway is not “do more schema.” It is “do schema for the right reason, and stop chasing the dead parts.”

In practice that means:

• Lean on the foundational entity types. Organization and LocalBusiness, Person, Service, Product, Breadcrumb, Article. These define what you are and how you connect, which is the part machines actually consume.
• Keep using structured Q&A even without the snippet. I still mark up FAQ content, including on my own site, not for a rich result Google no longer shows, but because a clean question-and-answer block is exactly what an answer engine lifts and quotes.
• Make it true. Google requires that your markup match what is visible on the page, and so do I, because the moment your schema and your content disagree you have taught a machine to distrust you.
• Keep it consistent everywhere. The facts in your markup should match your site, your profiles, and your listings. Conflicting data is how you lose a machine’s confidence, and schema is part of that data.

None of that is a growth hack. It is closer to hygiene, the unglamorous structural work that quietly decides whether a machine can recommend you without guessing.

So, does it still matter?

The question hides an assumption: that structured data was an SEO tactic for winning rich results. It was always really about making a real thing legible to machines, and for a while we happened to get snippets as a reward along the way.

Now the snippets are fading and the machines are writing the answers. Being legible to those machines is no longer a nice extra sitting on top of your SEO. It is the game. So yes, structured data still matters. It mostly just stopped being optional.

Frequently asked questions

Sources

• Google Search Central, General structured data guidelines (rich-result eligibility versus ranking)
• Google Search Central, How structured data markup works (understanding pages and entities)
• Google Search Central, The structured data features Google Search supports (the current supported list)
• Google Search Central, Local Business (LocalBusiness) structured data
• Schema.org, LocalBusiness