What is Structured Data?

Structured data is a way of describing a webpage's content in a format built for machines rather than humans. A visitor reads a page and infers, from layout and context, that a block of text is a business address, a customer review, or an answer to a specific question. A crawler or language model can't reliably make that same inference from raw HTML alone, headings, paragraphs, and divs don't carry meaning on their own. Structured data closes that gap by explicitly labeling content using a shared vocabulary, most commonly Schema.org, embedded in a page as JSON-LD.

Common schema types include LocalBusiness (name, address, phone, hours), Organization, Review and AggregateRating, Article, and Product. Each one tells a search engine exactly what a given piece of content is, not "here's some text that looks like a price," but "this specific text is the price." That precision is what lets Google render rich results such as star ratings directly in search results.

Two types no longer produce a Google search feature. HowTo rich results stopped showing on desktop and mobile in September 2023. Google ended FAQ rich results on 7 May 2026, and the Search Console report and Rich Results Test support are reported to follow in June 2026, with API support in August 2026. Markup you already have is harmless and need not be stripped.

Keep that statement scoped to what it covers. It says a Google SERP treatment went away. It does not say the data underneath stopped being legible to anything else that parses a page. Explicitly delimited question-and-answer pairs, or explicitly delimited procedure steps, are plausibly easier for an AI crawler to extract than the same content written as prose. That is a mechanism argument and an association, not a demonstrated effect, and a July 2026 survey of 45 studies in this area found no reviewed technique with a stable, cross-platform causal effect. Worth keeping on those terms; not worth claiming a return on.

The wider caution still holds. Google's May 2026 guidance states plainly that no special structured data is required for its generative AI features, so treat schema as a rich-result and entity-clarity measure first. Studies reporting that schema-marked pages get cited more are almost certainly measuring site quality, since sites that implement schema properly tend to do everything else properly as well.

Do not fold llms.txt into any of this. It is not a schema type, it never produced a search feature, and so it has no deprecation to reason around. It is an access convention aimed at AI crawlers. Google has said llms.txt is not needed for Google Search and has no effect on visibility or rankings (Google, June 2026). Other AI systems may read it, but support there is unconfirmed.

In an AI search context the case for structured data is a mechanism argument and should be held loosely. A system extracting an answer from a page has to resolve ambiguity, and unambiguous markup plausibly makes that easier, but that is an association rather than a demonstrated effect, and Google's guidance says nothing special is required. The more defensible use is entity clarity: schema is how a business tells search engines and AI models exactly who it is, which matters both for local search and for being represented accurately across the wider web (see GEO).

The Ritam Labs paid platform checks structured data as part of both SEO On-Page (D2) and AEO/GEO Readiness (D3), since the same markup that serves traditional rich results is plausibly easier for AI systems to extract from, though that is a mechanism argument rather than a demonstrated effect, and Local SEO (D6), where LocalBusiness schema specifically is checked alongside NAP consistency. Missing or malformed schema is among the findings a fix generator can resolve mechanically. In the free audit, structured data is checked in AI Search Readiness (FAQPage, LocalBusiness, HowTo, Speakable and JSON-LD coverage) and as LocalBusiness presence in Local SEO.