Most schema markup guides you will find in 2026 are recycled from 2018. They tell you to add HowTo markup for step-by-step rich results and FAQPage markup for expandable questions in the search results - advice that was correct seven years ago and is wrong now. Google has deprecated some of the most-recommended schema types, restricted others to a narrow set of sites, and quietly shifted the value of structured data toward something the old guides never mention: machine readability for AI systems.
This guide takes the current state of structured data seriously. It covers which schema types still earn rich results in Google, which ones Google has walked back, what structured data now does for AI search and large language models, and how to actually implement it without overcomplicating your stack.
What Schema Markup Is, Briefly
Schema markup is structured data: a standardized vocabulary, defined at schema.org, that you add to a page to describe its content in a machine-readable way. Instead of leaving a search engine or an AI system to infer that "$49" is a price, that "4.6" is a rating, or that a block of text is a recipe, schema states it explicitly.
The recommended format in 2026 is JSON-LD - a block of structured data placed in the page's HTML, separate from the visible content. Google has preferred JSON-LD for years, it is the easiest format to maintain, and the older inline formats (Microdata and RDFa) offer no advantage for new implementations. If you are starting fresh, use JSON-LD and do not look back.
The critical mental model: schema markup does not directly improve your rankings. Google has been explicit that structured data is not a ranking factor. What schema does is make a page eligible for enhanced presentation - rich results - and make its content unambiguous to the systems parsing it. Those are real benefits. "Higher rankings" is not one of them, and any guide that promises it is selling you something.
The Schema Types That Still Earn Rich Results in 2026
This is the part where the old guides go wrong. Here is the honest current state.
Product schema remains one of the highest-value structured data types. For e-commerce and any page selling a product, Product markup with price, availability, and review data can produce rich results showing price and rating directly in search. This still works and still matters. Product schema is the clearest "do this" on the list for anyone selling something.
Review and AggregateRating schema continues to power star ratings in eligible rich results, used within Product and a defined set of other types. Google has cracked down on self-serving review markup over the years - reviews must be genuine and the markup must reflect content actually on the page - but legitimate review data remains valuable.
Article schema is appropriate for news, blog, and editorial content. It does not produce a flashy rich result the way Product does, but it helps Google understand authorship, publication date, and content structure, and it is a sensible baseline for any publisher. It also feeds Google's Top Stories and Discover eligibility for qualifying sites.
Breadcrumb schema generates the breadcrumb trail shown in search results in place of a raw URL. It is low-effort, low-risk, and still works. Implement it.
Organization schema describes the entity behind the site - name, logo, contact points, social profiles, and identifiers. It does not create a rich result on its own, but it has become more important, not less, for reasons covered in the AI section below. Put accurate Organization schema on your site.
Event, Recipe, Video, and Job Posting schema continue to drive specialized rich results within their niches. If you publish that kind of content, the markup is worth implementing and remains supported.
The Schema Types Google Has Walked Back
Here is what the 2018-era guides will not tell you.
HowTo schema is dead as a rich result. Google fully deprecated HowTo rich results in September 2023 - they were removed across all devices and have not appeared in search since. If a guide is still telling you to add HowTo markup to win step-by-step rich results, that guide has not been updated in years. You can still use HowTo markup to describe content for machine readability, but the rich result is simply gone.
FAQPage rich results have been fully deprecated. This is the single biggest change from the old advice. Google first restricted FAQ rich results to authoritative government and health sites in 2023, and then removed them entirely - as of May 7, 2026, FAQ rich results no longer appear for any site, including the government and health sites that briefly kept them. For every commercial and editorial site, adding FAQPage schema produces no expandable question-and-answer treatment in search results at all. FAQPage was the most over-recommended schema type of the last decade, and that recommendation is now simply obsolete as a rich-result play.
That does not make FAQPage schema worthless - but the reason to use it has changed entirely, which brings us to the most important shift.
What Schema Now Does for AI Search
The forward-looking reason to care about structured data in 2026 is not the rich result. It is machine readability for AI systems.
AI search experiences - Google's AI Overviews, ChatGPT's browsing and search, Perplexity, and the rest - all work by parsing web content and synthesizing answers. Structured data makes a page's content explicit and unambiguous to exactly that kind of system. When your page states clearly, in structured form, what an entity is, what a product costs, who authored an article, and what question a passage answers, you remove the guesswork from the systems deciding whether to cite you.
This is why FAQPage schema is still worth implementing even though it no longer earns a Google rich result at all. A well-structured FAQPage block presents a clean question-and-answer pair that an AI system can lift and attribute directly. The same logic applies to Organization schema: as AI systems build their understanding of entities, accurate, structured information about your organization helps them represent you correctly rather than guessing.
Be honest about the state of the evidence here. No major AI platform has published a formal statement that structured data is a direct input to citation selection, and you should be skeptical of anyone who claims certainty. What is defensible is the underlying logic: structured data reduces ambiguity, and reduced ambiguity helps any machine - search crawler or language model - parse and represent your content accurately. Implement schema for that reason, not on the promise of a guaranteed AI ranking boost that nobody can actually deliver.
How to Generate and Implement Schema
You have three realistic paths.
Manual JSON-LD. Writing the JSON-LD by hand, guided by schema.org and Google's structured data documentation, gives you the most control and the cleanest output. For a small site or a developer-led team, this is very workable. The structure is not complicated once you have done it a few times.
CMS plugins and SEO platforms. On WordPress, the major SEO plugins generate schema automatically - RankMath has robust built-in schema support, Yoast generates a baseline schema graph, and dedicated tools like Schema Pro exist specifically for structured data. These handle the common types well and are the right choice for most non-developer teams. The trade-off is less control and occasional bloat - plugins sometimes emit schema for types you do not need.
Hand-built in the template. For custom-built sites, the cleanest approach is generating JSON-LD directly in the page templates from your actual content data, so the markup is always in sync with what is on the page.
Whichever path you choose, two rules hold. First, the markup must accurately reflect content that is actually visible on the page - schema describing content that is not there violates Google's guidelines and can earn a manual penalty. Second, validate everything. Use Google's Rich Results Test to confirm rich result eligibility and the Schema.org validator to check for syntax errors. Validate after every significant change.
The Practical 2026 Priority Order
If you are deciding where to spend limited time, this is the order that reflects current reality:
1. Organization schema sitewide - low effort, growing importance for entity understanding and AI representation. 2. Breadcrumb schema - low effort, still produces a real search-result improvement. 3. Product and Review schema - if you sell anything, this is the highest-value rich result still available. 4. Article schema on editorial content - sensible baseline for publishers. 5. FAQPage schema - not for the Google rich result anymore, but for clean, liftable Q&A that AI systems can parse and attribute. 6. Specialized types (Event, Recipe, Video, Job Posting) - only if you publish that content, and they remain genuinely valuable where they apply.
What is not on this list is as important as what is. Do not spend time on HowTo markup expecting a rich result. Do not assume FAQPage will give an ordinary site the old expandable treatment. And do not believe any source that tells you schema markup will directly raise your rankings. Implement structured data because it makes your content unambiguous to the machines that increasingly decide who gets seen - and judge it by that standard.