- What schema is (and what it isn’t)
- Why schema still matters when rich results come and go
- What “AI search” changes: extraction, grounding, and provenance
- The schema benefits that actually show up in real businesses
- What to prioritise (without turning schema into theatre)
- A quick note on schema for e-commerce
- The trap: schema as “marking up fantasies”
- The honest conclusion
I’ve been an advocate of schema for years. Not the “throw some markup on it and pray for rich snippets” version — the right use of schema: clarity, identity, and machine-readable truth.
Back in 2022, I went far enough down that rabbit hole to build my own schema generator, specifically to include entity markup properly — because most tools at the time treated schema like a formatting exercise, not an understanding problem.
Now that AI can spit out JSON-LD on demand, the “schema generator” part isn’t the clever bit anymore. The clever bit is the discipline behind it: defining entities properly, keeping identifiers stable, and making sure the markup reflects reality — not whatever was convenient to output.
That context matters, because the conversation around schema has changed. Search has shifted from “ten blue links with optional decorations” to a messy blend of classic results, AI-generated summaries, answer-style interfaces, and surfaces that pull information in ways most businesses don’t really understand. The old schema pitch (“get more SERP real estate”) isn’t just tired — it’s often wrong.
So what does schema actually do in 2026, when AI systems are doing more synthesis and less simple retrieval?
It does what it always did when used properly: it reduces ambiguity.
And that’s still valuable. Just not in the way most SEO blog posts promised.
What schema is (and what it isn’t)
Schema is a vocabulary for describing things — organisations, people, products, articles, events — and the relationships between them, in a way machines can read without guessing.
It is not a ranking lever you can “turn on”.
It is not a replacement for content quality, authority, internal linking, or basic technical competence.
And it is definitely not a way to “rank in ChatGPT”.
Google’s own documentation is blunt: structured data helps Google understand your content and can enable certain search features, but it doesn’t guarantee those features will show, and it doesn’t magically improve pages that don’t deserve to perform.
That’s the correct starting point. Everything else is details.
Why schema still matters when rich results come and go
If you were around for the FAQ markup gold rush, you’ll remember the pitch: add FAQPage schema, get big dropdowns in the SERP, dominate the results, improve CTR.
Google then restricted FAQ rich results so they show mainly for “well-known, authoritative” government and health sites, and not “regularly” for everyone else.
That change is annoying if you were treating schema as a SERP decoration strategy. But it’s useful as a lesson:
Rich result formats are temporary. Machine-readable clarity isn’t.
AI-shaped search surfaces don’t need your page to win a special snippet to benefit from structured clarity. They need to be able to extract, ground, and attribute information with fewer errors.
Schema is one of the cleaner ways to do that.
What “AI search” changes: extraction, grounding, and provenance
AI systems are not “reading” your page like a human. They’re extracting signals, matching entities, comparing sources, and synthesising.
In that environment, schema tends to help in three unglamorous ways:
Extraction reliability
When a system needs to answer, “What is this? Who wrote it? What product is being described? What’s the price? Where is the business located?” prose can be interpreted ten different ways.
Structured data is you saying: here are the fields, here are the values, here is what they mean.
This is why Product, Offer, Article, Organisation and Breadcrumb markup still matter. They make the “shape” of your content more legible.
Entity grounding (reducing confusion about who you are)
If your brand name is generic, your founder shares a name with three other people, or you operate in multiple markets, ambiguity kills you quietly. Not just in classic search, but in how AI systems represent you.
Schema properties like sameAs can help link your entity to stable identifiers (Wikipedia, Wikidata, official profiles), but only when you use them correctly. sameAs is identity, not “this sounds like a good reference.” Misuse it and you undermine trust.
You can add mentions for extra entity clarity, but treat it as optional garnish: it’s most useful on CreativeWork pages (articles, FAQs, guides) to list secondary entities referenced, as long as you’re consistent about what those entities are and you use stable @id identifiers (your own entity nodes, Wikidata, etc.).
Google has also said it can make “general use” of sameAs and other schema.org structured data — including for possible future features.
That’s not a promise, but it’s a clear signal that entity clarity is not wasted effort.
Provenance (who said it, and when)
AI answers increasingly lean on “is this credible and current?” questions, even if the user never asks them explicitly. That’s why author details, publisher identity, and dates matter.
Google explicitly points publishers toward datePublished and dateModified in structured data to help it understand dates correctly.
And Google’s guidance on Article structured data makes the point that it helps Google understand the page and show better title/image/date information across Google surfaces.
Again: not magic. But it reduces the chance your content gets misinterpreted or misattributed.
The schema benefits that actually show up in real businesses
Here’s what I see when schema is implemented properly (and maintained), without turning it into a hobby project:
Fewer misunderstandings about the business.
Correct Organisation/LocalBusiness markup helps disambiguate who you are, what you do, and where you operate. Google’s own Organisation guidance focuses on exactly that kind of clarity.
Cleaner “facts” in search surfaces.
Dates, authors, product attributes, breadcrumbs, and brand identifiers are less likely to be guessed incorrectly.
Eligibility for the rich results that still matter.
Not every rich result died. Some are still valuable when you genuinely qualify — especially in ecommerce and local contexts. But eligibility is conditional, and the content must match what you’re marking up.
Better internal discipline.
This is the underrated bit: schema forces organisations to decide what’s true. What’s the official name? What’s the canonical address? Who owns content? What’s the product catalogue structure? If you can’t answer these cleanly, your SEO problems aren’t “technical.” They’re organisational.
What to prioritise (without turning schema into theatre)
If you’re looking for schema that typically pays rent, it’s the foundational stuff:
Organisation / LocalBusiness for identity and disambiguation.
Article / BlogPosting for authorship and dates.
BreadcrumbList for hierarchy and context.
Product / Offer where you have real, user-visible product data.
Everything else depends on your business model. And if you’re doing it “because schema is good for SEO” without a specific outcome in mind (eligibility, disambiguation, extraction, governance), you’re probably doing busywork.
This is where an SEO sparring partner earns their keep: not by “adding schema”, but by pressure-testing what matters, what’s maintainable, and what will quietly break six months from now.
A quick note on schema for e-commerce
For e-commerce, schema is less about “looking nicer in Google” and more about making product information extractable without guesswork. Product and Offer markup (price, currency, availability, GTIN/SKU, brand, variants) gives machines the commercial facts; shipping and returns details make those facts usable for real purchase decisions.
The unglamorous part is the part that matters: accuracy, freshness, and consistency across your site, because outdated stock status and phantom prices train systems — human and machine — to distrust you. If you want a glimpse of where this is going, notice how much current guidance and tooling is shifting toward “merchant experiences” and structured commerce details rather than pretty snippets.
(We’ll do the full e-commerce/agent-shopping version properly later, because it deserves more than a paragraph.)
The trap: schema as “marking up fantasies”
A final warning, because it’s where a lot of schema implementations go wrong.
Schema is only helpful if it’s truthful and consistent.
If you mark up reviews that aren’t present, invent FAQs purely for markup, misuse sameAs to glue your brand to entities you don’t actually match, or apply structured data at scale without governance, you’re not improving understanding. You’re manufacturing confusion.
And in an AI-heavy search environment, confusion is expensive. Machines don’t just fail to reward you — they may learn not to trust you.
The honest conclusion
Schema in the age of AI search is not about hacks. It’s about being unambiguous.
It helps search systems — classic and AI-driven — understand what your pages are, what entities they refer to, and what facts you’re asserting, with less guesswork and fewer errors. It can still unlock rich result features where they exist, but that’s a side effect, not the strategy.
If your schema plan starts with “how do we get more SERP features”, you’re already behind.
If it starts with “how do we make our brand, content, and offers harder to misinterpret,” you’re thinking like someone who wants visibility that lasts.
That’s the difference between hype and utility.