Schema Markup Won't Get You Cited by AI. Google Says So.

A lot of agencies are currently selling schema markup as the way to get into AI Overviews. Google's own documentation says it isn't: "You don't need to create new machine readable files, AI text files, or markup to appear in these features."

That's from Google Search Central's page on AI features and your website. It's not a leak or an interpretation — it's the published guidance.

We implement structured data on every site we build. We just don't tell clients it's an AI visibility lever, because it isn't one.

What the data says

Beyond Google's statement, there's measurement. An Ahrefs study of 1,885 pages found that adding JSON-LD schema produced no meaningful lift in AI citations — in fact a small decline of around 4.6% on Google AI Overviews.

John Mueller of Google's Search Relations team made the adjacent point back in April 2025: structured data doesn't improve rankings, and it powers the specific search features listed in Google's structured data gallery. That's the job it does.

Two things can be true. Schema is worth implementing. Schema is not why anyone gets cited by an AI assistant. The industry has collapsed those into one claim because the first is easy to sell.

So why do we still implement it?

Three real reasons, none of which are AI:

Classic rich results. Product, Event and similar schema still drive measurable click-through improvements in ordinary search results. That's a direct, observable benefit.

Machine-readable clarity. Prices, ratings, dates, opening hours, service areas. Marking these up means systems interpret them correctly rather than guessing.

It forces content discipline. Genuinely — the exercise of marking a page up tends to expose that the page doesn't actually state its own facts clearly. The schema is downstream of a content fix that was worth making anyway.

None of those are "and then AI will recommend you."

What actually drives AI citations

The things that do correlate, from what's currently observable:

Answer-first structure. Roughly 44% of LLM citations come from the first 30% of a page. Put the answer at the top, in 40–60 words, before the context
Clean heading hierarchy. Around 68.7% of cited pages have strict, logical heading structure, against roughly 40% of uncited ones
Tables. Substantially more likely to be cited than the same information in prose
Named authors. Attributed content earns meaningfully more citations than unattributed
Freshness. Content refreshed within 30 days earns roughly 3.2x more citations than stale content
Third-party mentions. Around 85% of AI brand mentions originate from pages that aren't yours

That last one is the uncomfortable part, and it's why we take the research and software sections seriously. You cannot mark your way into being cited. You get cited when other people's pages say things about you — which means original data, genuine expertise, and being worth referencing.

The `llms.txt` question

Same answer. It's a proposed standard for telling AI crawlers about your site, and it's being sold as necessary.

No major AI system currently requires it. Google's documentation covers this in the same sentence about not needing new machine-readable files. If someone quotes you for implementing llms.txt as an AI visibility service, you're paying for a text file that nothing reads yet.

It costs nothing to add, and adding it is fine. Paying for it as a strategy is not.

What to do instead

Restructure your best pages to answer the question in the first paragraph, fix your heading hierarchy, put comparable data in tables, attribute everything to a real person, and refresh quarterly. Then work on being worth citing — which is a content and expertise problem, not a markup problem.

The full guide to AI search visibility covers the whole approach, including what else is being oversold. If you'd rather check where you currently stand, the AI visibility checker is free and doesn't ask for your email.

Sources

Written by Jared DeValk, founder of Nashville Digital. Published September 8, 2026.