Structured data for AI: where to start on WordPress

On WordPress, structured data for AI starts with an audit rather than an installation: your SEO plugin already publishes markup on every page, and the first real win is correcting your company identity inside that existing markup instead of stacking on another plugin.

This article covers what your site already outputs without you doing anything, what Google actually requires for its generative answers, the order in which to handle the rest, and the three situations where structured data is not your problem.

What is structured data, and what does it do for AI engines?

Structured data is a block of code invisible to your visitors, written in JSON-LD (JavaScript Object Notation for Linked Data), that states in machine language what a page contains: the exact company name, the address, the author of an article, the price of a product, the date of an event. The vocabulary comes from schema.org, a shared reference used by the major engines.

For answer engines, the value of that markup is the removal of doubt. An engine reading your homepage otherwise has to infer your legal name, your trade and your service area from a headline and a navigation menu. JSON-LD declares those facts instead of leaving them to be guessed. It gets no company cited, it lowers the odds that an AI gets you wrong.

Does Google require special markup to appear in its AI answers?

No, and Google says so in plain terms. In its documentation titled « AI features and your website », updated on December 10, 2025, Google states that a page only needs to be indexed and eligible to appear in Google Search with a snippet in order to be shown as a supporting link in AI Overviews or AI Mode. The same page adds that there are no additional technical requirements, and no special schema.org structured data you need to add for those features.

That sentence rules out a share of the offers being sold in 2026 around markup presented as « AI-specific ». There is no reserved vocabulary for generative engines that Google rewards, and a vendor selling you one is selling a category the engine's own publisher does not recognise.

The same sentence does not make structured data pointless. Google is describing eligibility for its own surfaces, not how ChatGPT, Perplexity or Claude interpret a page they read live, and not what a software agent does when it looks for your phone number or your opening hours. Markup stops being an entry ticket at Google, and remains a statement of identity readable everywhere else.

What does your WordPress site already output on its own?

Far more than most owners assume. WordPress ran 40.7% of all websites and 58.9% of those whose content management system is identified, according to the W3Techs reading of August 30, 2026. That scale has a direct consequence: the most widely used SEO plugins have industrialised markup, and ship it on by default.

The Yoast SEO developer documentation spells out what the plugin emits with no configuration at all: an Organization entity, a WebSite entity and a WebPage entity on every page, plus a BreadcrumbList trail and ImageObject records, with an Article entity added on every blog post. These pieces are not stacked side by side, they are cross-referenced into a single graph.

So the work that remains is almost never adding markup, it is filling it in. The graph is generated even when the fields are blank, approximate, or inherited from a trading name you dropped two years ago. A site can therefore declare cleanly, on every single page, an organisation with no logo, no address, and a name that no longer matches the sign above the door.

Where should you actually start, and in what order?

The useful order follows the commercial value of each declaration, not its technical difficulty. All four levels are set in admin forms, without writing a line of code.

• Your identity, first. Exact name, logo, official site address, third-party profiles. This is the block answer engines cross-check to decide which company they are talking about, and the one that does the most damage when it is wrong.
• Your premises, if customers come to you. Google makes only two properties mandatory for a local business, the name and the postal address, and recommends adding opening hours, telephone, geographic coordinates and price range.
• Your products or services, with a price and an availability a machine can read. This is the one place where markup changes a buying answer.
• Your long-form pages, last. Named author, genuine last-modified date, declared subject.

One rule outranks all four: the markup must say what the page says. Opening hours declared in JSON-LD that contradict the hours in your footer, a marked-up price that is no longer the one displayed, an outdated trading name, all of these are worth less than no markup at all. An engine that spots a contradiction does not pick the correct version, it loses confidence in both.

Why is FAQ markup no longer the priority you were sold?

Because Google no longer displays it. Google's own FAQPage documentation states that the FAQ rich result is no longer shown in Google Search results, a decision announced in its May 2026 changelog, and the corresponding documentation page was removed on June 15, 2026. FAQ no longer appears in Google's gallery of supported rich results.

This is a textbook case for a buyer, because FAQ markup was the single hardest-pushed selling point of WordPress SEO plugins for five years. What disappeared is a display decided by one engine, not the value of a page that asks a real question and answers it in three sentences. That content stays just as extractable by a generative AI as it was before.

The commercial lesson fits in one line: markup bought for a display can be discontinued by one engine's unilateral decision, whereas clear content stays readable by everyone who comes after. Set your budgets accordingly.

Should you go further than the plugin does?

Plugin vendors are now pushing past the individual page. Yoast announced on March 3, 2026 an opt-in consolidation feature that gathers a whole site's markup behind a single endpoint aimed at AI systems, rather than forcing them to crawl page by page to rebuild the links between an author, an article and an organisation.

The call for a small business stays simple. These options cost little once the plugin is installed and paid for, they earn their place on a site that publishes regularly, and they make up for no editorial gap whatsoever. Neatly consolidating the markup of a site with nothing to cite produces a well-formed statement about nothing.

When is structured data not your problem?

Three situations make this job secondary, and it pays to recognise them before committing a budget or a consultant's day.

• Your pages are unreachable by answer crawlers. Flawless markup behind a closed door is never read. Access gets settled before vocabulary.
• Your site is a three-page brochure with no substantive content. The markup will correctly describe a company about which there is nothing to cite, and the gap is editorial, not technical.
• Your real problem sits off-site. Answer engines cross-check several sources, and a company absent from trade directories, local press and customer reviews lacks raw material that no JSON-LD manufactures.

In all three cases the money is better spent on one genuinely useful page, on opening up access, or on showing up in the sources engines cross-check. Structured data amplifies existing material, it does not create any.

How do you check what your site actually declares?

A useful check answers three separate questions: is the markup present, is it valid, and does it tell the truth about your company? The first two are settled with a testing tool. The third means rereading your own declarations, because no machine knows that an address belongs to premises you left. The method Biencité applies, its three axes and nineteen signals, is published at https://www.biencite.fr/en/methode.

Two articles on the Biencité blog carry this further depending on your situation. « Shopify, Wix, Webflow, Framer: what your platform actually lets you do for AI » explains why WordPress leaves you freedom that other platforms lock away. « What an AI visibility scan actually measures, and what it cannot » sets out what an automated diagnostic can settle, and what it leaves to your judgement.

One last caveat applies to every markup offer you will come across. No structured data gets a company cited by an AI. It makes a company understandable; what makes it citable is what the company has written.

Want to know what your WordPress site declares to answer engines today? The diagnostic is free and read-only, with no account to create and nothing to install: https://www.biencite.fr/en/scan