Structured data for AI discoverability

Generative engine optimization has a hype problem.

Every week, someone invents a new checklist that promises AI visibility: add schema, add llms.txt, allow every crawler, block every crawler, write for embeddings, rewrite everything as Q&A, publish more lists, publish fewer lists, create entity maps, add author pages, add citations, add tools.

Some of those ideas are useful, but none of them guarantee that ChatGPT, Google AI Overviews, Perplexity, Claude, Gemini, or any other answer engine will recommend you. Structured data sits in that same category: it is useful, but it is not magic. Its real job is entity clarity.

It helps machines understand who is behind the site, what a page represents, who wrote a piece of content, which organization published it, and how important things relate to each other.

That matters for AI search because AI systems still need to retrieve, parse, compare, summarize, and cite information. But markup does not replace useful content, crawlability, authority, or user trust.

How AI systems encounter your content

AI systems can interact with websites in several ways.

Some use search indexes.

Some use their own crawlers.

Some fetch pages in response to a user action.

Some use licensed or partner data.

Some combine several methods.

OpenAI documents separate crawler/user-agent categories, including OAI-SearchBot for ChatGPT search features, GPTBot for content that may be used in training generative AI foundation models, and ChatGPT-User for certain user-triggered browsing actions.

Perplexity documents PerplexityBot as a crawler designed to surface and link websites in Perplexity search results, and Perplexity-User for user-requested actions.

Google documents Google-Extended as a robots.txt control token for managing whether content Google crawls may be used for Gemini model training and grounding in Gemini-related products. Google also states that Google-Extended does not affect inclusion or ranking in Google Search.

That is already more complicated than “allow AI bots.”

Robots controls matter, but they are not the whole story. Google’s own robots.txt documentation warns that robots.txt cannot enforce crawler behavior; it depends on crawlers obeying the rules.

So the practical approach is:

Make public content crawlable by the search systems you care about.
Decide crawler policies intentionally.
Keep important content available as text, not locked in images or scripts.
Use structured data to clarify entities and page purpose.
Avoid promises about AI placement.

Google’s AI features guidance is not hype

Google’s AI features documentation says there are no additional requirements to appear in AI Overviews or AI Mode, and no special schema.org structured data that you need to add.

That line should kill a lot of bad advice.

It does not mean structured data is useless. The same page also lists normal SEO fundamentals as worthwhile, including crawlability, internal links, textual content, page experience, and structured data that matches visible text.

That is the honest middle ground: there is no special “AI schema,” but there is still value in clear structured data. The best use of JSON-LD for AI discoverability is not to target an AI feature directly but to make the website’s entities and relationships less ambiguous.

Entity clarity: the types that actually help

For most sites, a small set of schema types matter more than a giant graph.

Organization

Organization schema on the homepage helps clarify the brand behind the site. It can connect the name, homepage, logo, and official profiles.

That matters because AI systems and search systems need to know whether “Acme,” “Acme Security,” and “Acme.io” are the same thing.

WebSite

WebSite schema can support site name clarity on the homepage. It helps distinguish the website from the organization that publishes it.

Article or BlogPosting

Article schema can clarify editorial content: headline, author, publisher, date, image, and canonical page identity.

This is useful when the URL is genuinely an article. It is not useful when every page is marked as an article by default.

Person or ProfilePage

Author pages can help connect content to real people or organizations. For content where expertise matters, a visible author page with structured data can reduce ambiguity.

Do not invent fake author entities. That is worse than having no author markup.

FAQPage

FAQPage can describe visible Q&A content. It is not a guaranteed rich-result or AI citation tactic. Its value is making question-and-answer structure explicit when that structure is genuinely on the page.

Callout: The AI-search value of structured data is not “add more schema.” It is “make the important entities easier to identify and connect.”

llms.txt vs JSON-LD

llms.txt and JSON-LD solve different problems.

The llms.txt proposal describes a Markdown file at /llms.txt that helps language models find useful website information at inference time. It is a curated guide: here are the docs, policies, product pages, APIs, or reference materials that matter.

JSON-LD is page-level structured data. It describes the content and entities on a specific page.

They are complementary, not interchangeable.

A good /llms.txt can point an AI system toward high-value resources. Good JSON-LD can clarify what those resources are once fetched.

But neither one guarantees inclusion in AI answers.

Also be honest about adoption. llms.txt is a proposal, not a universal standard with guaranteed support from every major AI system. It may still be worth implementing because it is low-cost and useful as a curated machine-readable map, but it should not be oversold.

What not to promise

Do not promise:

“Add schema and ChatGPT will recommend you.”
“FAQ schema gets you into AI Overviews.”
“llms.txt guarantees AI visibility.”
“Blocking or allowing one bot controls all AI usage.”
“More structured data means more authority.”
“AI search has replaced SEO fundamentals.”

These claims are either false, unproven, or too broad to be useful.

Promise practical improvements instead:

The site identity becomes clearer.
Key pages become easier to classify.
Author and publisher relationships become more explicit.
Important content is easier to find through internal links.
Crawlers can detect gaps and inconsistencies.
The site avoids hidden or misleading markup.

That is less flashy, but it is real.

The crawl + GSC connection

AI discoverability work should not be separated from normal SEO data.

Google Search Console tells you which queries already expose your site. It shows question patterns, page impressions, click trends, and pages that are close to earning more visibility.

A crawl tells you whether those pages are technically understandable: titles, headings, canonicals, structured data, internal links, visible content, indexability, and entity signals.

Together, they can produce better recommendations.

For example:

A homepage has strong branded impressions but no Organization schema.
A blog post gets question queries but has no visible FAQ or short answer section.
An article ranks but has no author page or author URL.
A documentation page is important but buried with weak internal links.
Multiple pages mention the same product but use different names.

Those are practical AI discoverability opportunities because they improve clarity.

A sensible workflow

Start with the public entity.

Who are you? What is the canonical homepage? What profiles prove it? What logo represents the brand?

Then map the content types.

Which URLs are articles? Which are docs? Which are products? Which are support pages? Which pages answer high-intent questions?

Then add structured data only where it matches the visible page.

Then review crawler access.

Do not blindly allow or block everything. Decide what you want included in search-like AI features, what you want excluded from training where controls exist, and what should not be public at all.

Then create or improve /llms.txt if you have useful resources to curate.

Then monitor.

There is no one-time AI SEO switch, only the ongoing work of making the site easier to understand, easier to crawl, and harder to misinterpret. That is what structured data is good at.

Frequently asked questions

Does structured data help AI search?

It can help clarify entities and page purpose, but it does not guarantee AI citations, recommendations, or inclusion in AI answers.

Is there special schema for AI Overviews?

Google says there are no additional requirements and no special schema.org structured data needed for AI Overviews or AI Mode.

Should I add llms.txt or JSON-LD first?

They solve different problems. JSON-LD describes individual pages. llms.txt is a proposed way to provide a curated guide to important site resources.

Should I allow AI crawlers?

That depends on your goals. Some crawlers relate to search visibility, some to training, and some to user-triggered actions. Review each crawler’s documentation before deciding.

Explore the AI discoverability hub, AI discoverability overview, JSON-LD basics, and Organization schema.