Does schema markup still matter for AI search?
Yes, but not the way it is usually sold. Schema markup makes it easier for a machine to work out what a page is, who published it and how its parts relate, which helps an engine resolve your business and parse your content reliably. It cannot create an answer. Marking up a page that answers nothing produces a very well described nothing, which is why schema belongs near the end of the job rather than at the start of the invoice.
By Liron Segev. Last updated August 2026.
What it does and what it cannot do
What it does. It removes ambiguity. Instead of a machine inferring that a block of text is an author name, a publish date or a question and answer pair, you state it. That saves the engine work and, more importantly, removes the chance of it inferring wrongly.
What it cannot do. It cannot make a page relevant, make a claim credible, or answer a question the page does not answer. Structured data is a description of content. If the content is not worth citing, a perfect description of it is not worth citing either.
The reason this needs stating is that markup is easy to sell. It is technical, it is invisible to the client, it produces a tidy before and after in a validator, and it can be delivered in an afternoon. That combination makes it a popular deliverable and a poor centrepiece.
Which types are actually worth the effort
| Type | What it establishes | Verdict |
|---|---|---|
| Organization and Person | Establishes who publishes the site and who stands behind the content | Highest value. This is the entity work. |
| Article | Marks a page as a written piece with an author and dates | Worth it, mainly for the author link |
| FAQPage | Marks question and answer pairs explicitly | Cheap and useful, if the Q&A is real |
| BreadcrumbList | Describes where the page sits in the site | Small but free |
| Product, LocalBusiness, Event | Describes specific things with structured attributes | Valuable when they genuinely apply |
Your site probably has six of you
This is the one worth acting on, and it is the one plugins get wrong by design.
Most sites declare the author on each page as a fresh, self-contained block. Page one says the author is a Person named Liron Segev. Page two says the same. So does page three. To a human that is obviously one person. To something consuming the data it is three separate Person entities that happen to share a name, and the credibility that should accumulate to one identity gets divided between them.
The fix is to define the entity once, give it an identifier, and reference that identifier everywhere else instead of repeating the definition. Every article then points at the same author and the same publisher, and there is one thing accumulating authority rather than a dozen fragments of one.
This site had exactly that problem until August 2026. Our highest-value page, the visibility study, declared its own author and publisher inline while every other page referenced the shared identity. The single most cited page on the site was crediting a different entity than the rest of it. It validated cleanly the entire time, because this is not a syntax error. It is a modelling error, and no validator will tell you about it.
What is actually worth implementing
For most business sites the useful version of this is small, and it is finished in a day. The temptation is to keep going well past the point of return.
One shared identity, defined once. A single Organization describing the business and a single Person for whoever writes, each given a stable identifier, declared in one place that every page inherits. This is the part that matters most and the part almost nobody does correctly.
Article on every content page, referencing that identity. Author and publisher as references to the shared identifiers rather than fresh definitions. Real publish and modified dates, meaning the dates the content actually changed and not the dates the site was rebuilt.
FAQPage where genuine question and answer pairs exist. Only where the questions are visible on the page and the answers are real. This is cheap, it maps directly onto how buyers phrase things, and it is the one type most sites could add today with no downside.
Breadcrumbs that describe the real structure. If a page sits two levels down, say so. If your site is flat, do not invent a hierarchy to look organised.
That is the whole list for most sites. Add LocalBusiness if you have a physical location, Product if you sell discrete products, Event if you run events. Beyond that you are decorating, and the time is better spent on the questions your pages have not answered yet.
Four other ways this goes wrong
Marking up content that is not on the page. FAQPage schema containing questions a visitor cannot see is a misrepresentation, and it is treated as one.
Stale dates. A dateModified that updates every time the page is rebuilt claims freshness that does not exist. It is a small lie told automatically, thousands of times.
Stretching types to fit. Marking a services page as a Product because Product has more fields does not gain you anything and creates a description that contradicts the page.
Trusting the plugin. Read the JSON-LD your site actually serves. Settings screens describe intent; the served page is the only source of truth, and the gap between them is where most of these problems live.
Do this last, not first
The order that works: find the questions, write answers that survive being quoted, publish under a consistent identity, then mark it all up so machines do not have to guess. Schema at the end of that sequence compounds the work. Schema at the start of it describes an empty room in great detail.
If a proposal leads with structured data as the main deliverable, that is worth questioning. It is an afternoon of work presented as a strategy, and it is the easiest part of the whole discipline to deliver.
Related questions
Does schema markup help with AI search?
Yes, but indirectly and less than it is usually sold as doing. Schema makes it easier for a machine to establish what a page is, who published it and how its parts relate, which helps an engine resolve your business to one entity and parse your content reliably. What it cannot do is create an answer. Marking up a page that answers nothing produces well-described nothing.
Which schema types matter most for AI?
Organization and Person, by some distance, because they establish the entity behind everything else. Article and FAQPage are worth adding on content pages, mainly for the author connection and for making question and answer pairs explicit. BreadcrumbList is small but free. Product, LocalBusiness and Event are valuable where they genuinely apply and harmful where they are stretched to fit.
What is the most common schema mistake?
Redefining the same entity on every page instead of referencing it once. If each page declares its own Person node for the same author, a consumer sees many separate authors rather than one, and the authority that should accumulate to a single entity gets divided across pages. The fix is to define the entity once with an @id and reference that @id everywhere else.
Will schema markup alone get me into AI answers?
No. Schema describes content, it does not substitute for it. A page with no clear answer to a real buyer question will not be quoted regardless of how thoroughly it is marked up, because there is nothing worth quoting. Schema is a multiplier on content that already deserves to be cited and a multiplier of zero when it is not.
Do I need a plugin or a developer for this?
For most sites a plugin covers the basics adequately, and the basics are most of the benefit. Where plugins tend to fall short is entity consolidation, because they generate self-contained blocks per page rather than referencing one shared identity. That specific fix usually needs someone editing templates directly, and it is the fix worth paying for.
How do I check my schema is working?
Fetch the page and read the JSON-LD it actually serves rather than trusting the plugin settings. Confirm it parses, confirm the author and publisher resolve to a single shared identifier across pages, and confirm nothing describes content that is not on the page. Validators catch syntax errors; they do not catch a site that has quietly declared six different versions of the same author.
Other questions buyers ask
Find out where you actually show up
The free scan runs your business against ChatGPT, Claude, Gemini and Perplexity and shows you which buyer questions you appear in and which ones a competitor owns.