Structured data repeats information already written on a page in a machine-readable form; it adds no new information and doesn’t by itself get a brand into AI answers. Google states clearly that no special markup is needed for its AI features, and it removed FAQ rich results entirely in May 2026. The most valuable technical step is Organization markup that identifies the brand as a single entity. The most reliable signal is still information written clearly in page text.
When AI visibility comes up, one recommendation is heard often: “Add FAQPage schema to your site so LLMs read you.” It’s simple, easy to implement and feels like concrete work. That’s why it spreads.
The real picture is messier, and it changed noticeably in 2026. Some effects of structured data are well documented; some are guesses; some no longer apply. This article tries to separate which is which.
What does structured data do?
Schema.org markup repeats information on the page in a format machines can read easily: this is an organization, this is its name, this is its address; this is a product, this is its price; this article was written by this person. Search engines have used this markup for a long time and show rich results for some types.
The key point: structured data doesn’t add new information. It makes information already on the page easier to read. Google’s guidelines require markup to match visible page content; supplying information via Schema that isn’t on the page goes against the guidelines and weakens trust.
What happened to FAQ rich results?
It happened in two steps. In August 2023, Google limited FAQ rich results to “well-known, authoritative government and health websites.” On May 7, 2026, it added a note to its documentation that FAQ rich results no longer appear in Google Search. In the same announcement, Google said the FAQ report and Rich Results Test support in Search Console would be removed in June 2026, and Search Console API support in August 2026.
Google’s note also says FAQPage markup can stay on the page; FAQPage remains a valid Schema.org type. The markup does no harm, but it no longer has a visual payoff in search.
That doesn’t make FAQ content worthless. Short paragraphs that directly answer a clear question are exactly the kind of text AI summaries can quote. What’s valuable is the FAQ section itself; the markup is a secondary layer.
What does Google say about its AI features?
Google Search Central’s “AI features and your website” guidance is clear: there are no additional requirements, no special Schema type and no separate index for appearing in AI Overviews and AI Mode. SEO fundamentals apply: allow crawling, make content findable through internal links, present important information as text and keep structured data consistent with visible content.
Within that framework, structured data is a supporting signal: it can help Google understand content and organizations correctly, but it doesn’t guarantee inclusion in AI Overviews.
Do other LLMs read Schema?
The honest answer: it isn’t fully known. Two situations need to be separated.
Search-grounded answers. In systems that rely on a search index, such as Google’s AI features, the meaning a search engine extracts from structured data can carry into the answer indirectly. Especially for organization and product information, correct markup helps the brand be recognized correctly.
The model’s own knowledge and direct crawling. There’s no clear public statement that large language model providers give special weight to JSON-LD markup. Crawlers may see the markup when they read a page’s HTML, but how much it affects the answer is unclear.
Given this uncertainty, the sensible approach is: do the structured data, but don’t build your visibility strategy on it.
Which markup types come first?
| Type | What it contributes | Priority |
|---|---|---|
| Organization (with sameAs) | Recognizing the brand as a single entity; linking official profiles | High |
| Product / Service | A clear definition of product and service attributes | High (for brands selling products) |
| Article + Person (author) | Linking content to its author and their expertise | Medium |
| BreadcrumbList | Understanding site structure | Low |
| FAQPage | Clarifying FAQ structure for machines; no visual payoff in search | Low |
The sameAs field in Organization markup is especially valuable. It links the brand’s LinkedIn, directory and other official profiles to a single entity. It’s one of the few technical steps that strengthens the entity recognition link in the LLM visibility chain.
Does llms.txt actually work?
llms.txt is a file format proposed on September 3, 2024 by Jeremy Howard, co-founder of Answer.AI: a file in the site’s root that offers language models a summary of the site and markdown versions of key pages. Its main goal was quick access to software documentation for coding assistants.
Google’s position is clear: the Google Search team has said it doesn’t use llms.txt, including for AI Overviews, and John Mueller compared the file to the old “keywords” meta tag. Companies such as Stripe, Vercel, Cloudflare and Anthropic publish the file, but they do so for the coding agents developers use. For a B2B brand, no measurable contribution to AI visibility has been shown today. It can be added because the cost is low; there isn’t enough reason to put it high on the priority list.
What is the real signal?
In AI visibility, the most reliable signal is text clearly written on the page that directly answers the question. If certification scope, product capacity or a service’s pricing model is written in page text, the model has a good chance of finding it, with or without markup. If it isn’t on the page, markup can’t close that gap.
What we see in the field points the same way. Only a very small share of visibility gaps come from missing markup; most come from Content, Communication and Authority GAPs.
How does Recro handle structured data?
The Recro Insight Model checks technical signals at the accessibility and entity recognition links of the chain. A structured data recommendation enters the action plan only when a concrete gap is found at those links. Priority always goes to missing information in page text.
One simulation · 4 key pillars · 90+ steps
What sets Recro apart is four pillars built from scratch for each brand. Together they form a single simulation of 90+ steps, with each pillar producing the input for the next.
- Persona builder: Decision-maker profiles that represent the brand’s real buyers, specific to its sector and sales structure.
- Question builder: The brand-specific decision questions these people ask AI along their buying journey.
- Report builder: A report that reads the answers against the brand’s goals and competitors, together with the sources behind them.
- Action recommendation builder: A prioritized list of actions that closes the sector-specific signal gaps.
On technical signals, the difference is the starting point: Recro starts from “which information fails to reach the answer in the buyer’s question?”, not “which markup is missing?”
- The report builder extracts the links that source the conversation: are the brand’s pages cited, and if so, for which information? If a page is cited but the brand is misidentified, entity markup comes onto the agenda.
- The action recommendation builder puts structured data in its proper place in the sector-specific LLM signal list: high-return steps like Organization and sameAs move up; markup that makes no difference on its own doesn’t jump ahead of content actions.
Executive summary
- Structured data makes information on the page readable for machines; it doesn’t add new information. Google says no special markup is needed for its AI features.
- FAQ rich results were fully removed on May 7, 2026. FAQ content keeps its value; the markup is secondary.
- Priority order: information in page text, Organization and sameAs, Product, author markup. llms.txt isn’t a priority today.
Frequently asked questions
Should we remove FAQPage markup?
No need. Google says the markup can stay, and it clarifies page structure. But since it no longer produces rich results in search, it isn’t worth new investment.
Which markup type matters most?
For most B2B brands, Organization markup with the sameAs field. It links the brand’s profiles across platforms to a single entity.
Is adding structured data with plugins such as Rank Math enough?
For the basic types, usually yes. Check that the markup the plugin generates is consistent with the information on the page and that Organization details are complete.
Should we add an llms.txt file?
It can be added because the cost is low. But Google says clearly that it doesn’t use it, and its effect in other AI tools isn’t proven. It shouldn’t be high on the priority list.
How do we measure the effect of structured data?
Measuring it directly is hard. An indirect way is to run entity recognition questions (“What is [Brand], what does it do?”) repeatedly before and after adding the markup and compare the accuracy of the answers.
Is your visibility gap in the markup, or in information missing from the page? Recro’s brand-specific simulation measures technical and content signals together and puts priority in the right place.
Request a demo insight report →
Sources
- Google Search Central · AI features and your website
- Search Engine Journal · Google drops FAQ rich results from search (2026)
- Answer.AI · /llms.txt: a proposal to provide information to help LLMs use websites (2024)
- Search Engine Journal · Google’s llms.txt guidance
- Schema.org · sameAs
This article was prepared with AI assistance and published under the review of Mehmet Semih İpek.
