Schema Markup for AI Search: Which Types Matter in 2026
Start with the thing most articles on this topic won't tell you: Google has never published a schema requirement for AI Overviews. There is no AI-specific markup, and structured data is not a prerequisite for being cited.
Which makes the honest question narrower and more useful. Not "what schema gets me into AI answers," but "what does structured data actually do for machine readability, and which types still earn their implementation cost in 2026."
At a Glance
- No schema type is required for AI Overviews or AI Mode. Anyone selling you one is overstating it.
- What structured data does reliably is remove ambiguity: who you are, who wrote this, what it costs, when it was updated.
- FAQ rich results were fully retired in May 2026. HowTo rich results went the same way. Both types remain valid markup and still get parsed.
- Organization schema is the highest-value type for most B2B sites, because it connects your brand to consistent facts.
- The rule that matters more than type selection: markup must describe what a visitor can actually see.
- Implement on one page, verify, then template it. Site-wide rollouts before validation are how you get hundreds of pages with the same error.
What structured data actually does
Plain text is ambiguous to a machine. A page can mention a price, a company and a person without making clear which price belongs to which product, whether the company is the publisher or a competitor being discussed, or whether the person is the author.
JSON-LD removes that ambiguity. It states facts in a format that doesn't require interpretation, and it links entities to identifiers so systems can connect what's on your site to what they already know about you.
That's genuinely useful. It's also more modest than the claims usually made for it. Structured data helps a system understand your page. It doesn't make the page worth citing. Google's introduction to structured data is the primary reference, with type definitions maintained at Schema.org.
Which types still matter
The landscape changed in 2026, and several guides haven't caught up.
| Type | What it clarifies | Rich result status | Effort |
|---|---|---|---|
| Organization | Brand identity, logo, social profiles, contact | Knowledge panel eligible | Low |
| Article / BlogPosting | Author, publication and update dates | Supported | Low |
| Product / Offer | Price, currency, availability, SKU | Supported | Medium |
| Review / AggregateRating | Ratings tied to specific items | Supported | Medium |
| LocalBusiness | Physical location, hours, service area | Supported | Low |
| FAQPage | Question and answer pairs | Retired May 2026 | Low |
| HowTo | Discrete procedural steps | Retired | Medium |
Organization schema is where to start for most B2B SaaS. It's the type that connects your brand name to a consistent set of facts, and inconsistency there is a common reason systems get your company details wrong. Name, logo, social profiles, contact details, applied in your site template.
Article schema matters more than its rich result value suggests. Author attribution and update dates are signals about whether content is current and who stands behind it.
FAQPage and HowTo are the interesting case. Google restricted FAQ rich results in August 2023 to authoritative government and health sites, then retired the feature entirely in May 2026. Search Console reporting and Rich Results Test support follow.
Both remain valid Schema.org types. Google's documentation states you don't need to remove existing markup, and unused structured data doesn't cause problems. It's still parsed, and it's still readable by other crawlers.
The practical read: don't implement FAQPage expecting a visible search feature, because there isn't one anymore. Implement it if you want question-and-answer pairs stated unambiguously for machines. That's a smaller reason, and it's an honest one.
The rule that matters most
Markup must describe what a visitor can actually see on the page.
If your JSON-LD states a price that isn't displayed, an author who isn't credited, or a rating that isn't shown, that's a structured data policy violation. The usual consequence is Google ignoring the markup. The more serious consequence is a manual action.
Either way, you've spent implementation effort on something that either does nothing or actively hurts. This is the single most common structured data mistake and the easiest to avoid.
The fix sequence when you find a mismatch: correct the visible page first, then redeploy the markup, then validate. Not the other way round.
Implementing it
Pick one page. Something with real traffic or commercial value. Identify the facts you want stated unambiguously — company identity, author, price, dates.
Generate the JSON-LD. CMS plugins handle this for WordPress and Shopify without developer time. Custom sites need a template pulling from your database, which is more work upfront and far more reliable afterwards.
Include identifiers. SKUs for products, consistent author URLs for people. Identifiers are what let systems connect facts across your site rather than treating each page in isolation.
Deploy to staging and validate before production. Then check Search Console for new errors after it goes live.
Then template it. The point of getting one page right first is that you're about to replicate whatever you built across hundreds. Errors template just as efficiently as correct implementations.
Governance
Three roles, and they shouldn't collapse into one person.
The content owner signs off that the visible page states the facts correctly. The developer implements the markup. The reviewer validates and checks Search Console after deployment.
Then a quarterly audit for drift. Prices change, authors leave, offers expire, and markup quietly goes stale while nobody notices because nothing visibly breaks. Weekly Search Console error alerts catch the technical failures; only an audit catches the factual ones.
Measuring it
Be realistic about what's measurable.
Search Console shows you whether the markup is valid and whether rich results are appearing. That's the technical layer and it's the only part with clean data.
For AI citation, run a before-and-after prompt test. Ten questions your buyers actually ask, run against ChatGPT, Gemini and Perplexity before you implement and again four weeks later. Note whether you're cited and what facts get quoted.
The honest caveat: schema is one variable among many and you won't isolate its effect. If citations improve, structured data was probably part of it. Anyone promising you a measurable schema-to-citation attribution is selling something.
For the broader measurement framework, answer engine optimization covers citation share and how to track it properly.
Where this fits in AEO
Schema is a supporting practice, not a strategy. It removes ambiguity about facts you've already stated clearly. It doesn't make unclear content clear, and it doesn't substitute for the things that actually drive citation: extractable structure, factual consistency across your own pages, and corroboration from sources other than you.
Do it because it's cheap and it removes a category of confusion. Don't do it expecting it to be the lever.
Strivelabs samples ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews against prompt sets you define, which is how you'd measure whether any of this work is landing. It doesn't implement schema, that stays a developer job.
Track whether your structured data work is landing in AI answers
Strivelabs samples ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews against prompt sets you define.
Frequently Asked Questions
Is schema markup required for AI Overviews?
No. Google has published no AI-specific structured data requirement, and schema is not a prerequisite for appearing in AI-generated answers. It helps machines parse your facts unambiguously, which is a narrower and more defensible claim.
Should I still use FAQPage schema?
Only for machine readability. FAQ rich results were retired in Google Search in May 2026, with Search Console and Rich Results Test support removed shortly after. The markup remains valid and is still parsed, but there's no visible search feature to gain.
What happens if my markup doesn't match my visible content?
It's a structured data policy violation. Usually the markup gets ignored; in worse cases it results in a manual action. Either way you've wasted the implementation. Fix the visible page first, then the markup.
Which schema type should a B2B SaaS site start with?
Organization, applied site-wide in your template. It's the type that connects your brand to consistent facts about who you are, and it's the cheapest to implement.
Which format should I use?
JSON-LD. It sits in the document head rather than being woven through your HTML, which makes it far less likely to break during a redesign. One validated block per page.
Does schema improve rankings?
Not directly. It can make a page eligible for rich results, and it clarifies facts for any system parsing the page. Neither is a ranking factor in itself.
How often should I audit it?
Quarterly for factual drift, with weekly Search Console alerts for technical errors. Stale prices and departed authors don't throw validation errors, which is exactly why they persist.
Related Posts
Generative Engine Optimization: The Complete Definition
Your rankings can hold steady while this goes badly. What generative engine optimization is, how retrieval actually works, and who owns which part.
LLM SEO: How to Get Cited by ChatGPT and Perplexity
OpenAI runs three crawlers, not one, controlled separately. Block the wrong one and you close the route you wanted. The two ways into an AI answer, explained.
What Is an SEO Agent? Definition and Use Cases
A tool does what you ask. An agent decides what needs doing. If you have to notice the problem first, it's a tool. What SEO agents do, and where they fail.