Schema Markup for AI Overviews

Does schema markup for AI Overviews actually get a page cited? Is FAQ schema still worth adding, and which JSON-LD types matter now that Google answers so many queries on the results page itself?

These are fair questions, because structured data is the one part of SEO that looks like it was written for machines. The trouble is that the effort usually goes to the wrong place. Duplicate graphs pile up, retired markup stays on every post, and the things that actually decide eligibility (a stray noindex, a snippet restriction, a site-level setting in Search Console) never get checked.

The short answer to the first question is no. Google’s own documentation on AI features says there’s no special schema.org markup to add, and I take that at face value. I’d use schema to describe the page accurately and to earn rich results, and I’d judge AI Overview visibility by eligibility checks and Search Console data.

Below is what Google documents today, what I’d implement on a WordPress site and how to measure the result without fooling yourself.

Does Schema Markup Help AI Overviews?

Not directly. Google’s AI features documentation says there are no additional technical requirements for appearing in AI Overviews or AI Mode, and no special schema.org structured data to add. A page needs to be indexed and eligible to show a snippet in Search.

Google’s newer guide to optimizing for generative AI features goes further. It says structured data “isn’t required for generative AI search” and lists overfocusing on structured data among the things site owners don’t need to worry about, next to llms.txt files, content “chunking” and rewriting pages for AI.

Schema still has 3 jobs worth doing:

• Rich result eligibility. Breadcrumbs, product details, review stars and other supported features still read structured data.
• Clear facts about the page. The headline, author, dates and publisher become explicit instead of guessed.
• Consistent entities. One Organization or Person node, reused across the site, tells Google the same story on every URL.

Those are good reasons on their own. I’d just stop selling them, to a client or to yourself, as an AI Overview tactic.

How AI Overviews Choose Supporting Links

AI Overviews show a short summary with links on queries where Google thinks a summary adds value, and they don’t appear on every query. AI Mode handles longer, multi-step questions and follow-ups. Both are built on Search, so the links they show come from Google’s index and its ranking systems.

Two details explain why the links vary so much:

• Query fan-out. Both features can run several related searches in the background and pull supporting pages from each one, so a page can appear for a query it never targeted.
• Different models. Google says AI Overviews and AI Mode may use different models and techniques, so the same question can surface different links in each.

The 4th check is newer than most advice on this topic. Google’s generative AI guide now says a site must be included in Search generative AI features in Search Console to be eligible. Include is the default, so most sites are already in.

The Search Generative AI Setting

The setting lives under Settings in Search Console. It has 2 states:

• Include (default): links and content from the site can appear in AI Overviews, AI Mode and generative AI features in Discover.
• Exclude: the site’s links and content don’t appear in those features, and the site gets no impressions or traffic from them.

Google says this control isn’t used as a ranking or inclusion signal for the rest of Search. It doesn’t affect AI training either, which is what Google-Extended is for. A change generally takes a few days to apply.

Schema.org, Rich Results and AI Overviews

Most of the confusion comes from treating 3 different things as one. Schema.org is a vocabulary, and Google rich results are search features that read a small part of it. AI Overviews are a separate feature that needs neither.

Layer	What it is	Needs structured data?	What success looks like	How to check
Schema.org vocabulary	An open set of types and properties used by many systems	It is the markup	A valid, consistent graph	Schema Markup Validator
Google rich results	Search features with their own documentation	Yes, following each feature’s guide	An enhanced result, which Google never guarantees	Rich Results Test and Search Console reports
AI Overviews and AI Mode	Generated answers with supporting links	No	Impressions as a supporting link	Generative AI performance report

The practical consequence is that a type existing on schema.org doesn’t make it a type Google uses. And passing the Rich Results Test says nothing about whether an AI Overview will link to the page, because each tool checks only its own layer.

Controls That Change Eligibility

If a page never shows up in AI features, I’d look at these controls before touching any markup. Each one does something different, and 2 of them get mixed up all the time.

Control	Where it lives	What it changes	Effect on AI Overviews
robots.txt for Googlebot	robots.txt file	Whether Google can crawl the URL	A page Google can’t crawl can’t be a supporting link
noindex	Robots meta tag or X-Robots-Tag header	Removes the page from Search	Removes it from AI features too
nosnippet	Robots meta tag or header	No text snippet for the page	Text from the page isn’t shown in AI features
max-snippet	Robots meta tag or header	Caps the snippet length	Limits how much text Google can show
data-nosnippet	HTML attribute on an element	Hides that element from snippets	Only that text is held back
Search generative AI setting	Search Console, Settings	Site-level include or exclude	Exclude removes the site from AI Overviews, AI Mode and Discover AI features
Google-Extended	robots.txt user agent token	Use of content for Gemini training and grounding	None on Search or AI Overviews

Every tag in this table works only after Google recrawls the page, which can take days or longer. URL Inspection shows the HTML Googlebot actually fetched, so it’s the place to confirm a tag is really there.

Google-Extended Is Not an AI Overviews Switch

Google-Extended is a robots.txt token that controls whether Google can use crawled content to train future Gemini models and to ground answers in Gemini Apps and Vertex AI. The Google crawler documentation says it doesn’t affect a site’s inclusion in Google Search and isn’t a ranking signal.

So blocking it is a training decision. If the goal is to stay out of AI Overviews, the Search generative AI setting is the documented route, and noindex or nosnippet are the blunt ones. I covered the Cloudflare side in how to block AI training without blocking Google Search.

Which Schema Types Still Matter

For a blog or publisher site, the list is short. I rank types by how well Google documents them, how often a publisher needs them and how much damage a mistake does.

Priority	Type	What Google uses it for	Worth adding?
First	Article or BlogPosting	Headline, dates, images and author	Yes, on every post
First	Organization or Person	Site owner details, logo, knowledge panel facts	Yes, once, on the home or about page
Second	BreadcrumbList	The breadcrumb trail in results	Yes, when breadcrumbs are visible
Second	Product, Offer and merchant listings	Product snippets and shopping results	Only on real product pages
Second	Review and AggregateRating	Review snippets for supported types	Only for genuine reviews of one item
Retire	FAQPage	Nothing in Google Search since May 7, 2026	No, as a rich result or an AI tactic
Retire	HowTo	Nothing, since 2023	No
Skip	Types a plugin adds by default, such as a decorative ItemList	Nothing visible	No

Article and BlogPosting

Google’s Article structured data has no required properties. Google asks for the properties that apply, and these are the ones I’d always fill:

• headline that matches the visible title
• image with crawlable URLs, ideally in 16×9, 4×3 and 1×1 versions of at least 50,000 pixels each
• datePublished and dateModified in ISO 8601 format with a time zone
• author as a Person or Organization with name and a url or sameAs

Each author goes in their own author field. Google asks site owners not to merge several names into one string or put a job title inside name.

It is wise to pick the most specific type that’s true. BlogPosting fits a blog post, and NewsArticle suits news reporting. dateModified should change only when the content changed, because a plugin that bumps it on every save tells Google the page is fresher than it is.

Organization and Person

Google recommends Organization markup on the home page or one page that describes the organization. It doesn’t need to be on every page. For a personal site like this one, a Person node does the same job.

The fields I’d include:

• name, url and logo
• sameAs links to real profiles only
• contact details or an address only where they’re real and relevant
• the most specific accurate subtype, such as OnlineStore
• a stable @id, such as https://example.com/#organization

The @id matters more than people think. It lets every post point to the same entity instead of describing it again with slightly different details.

BreadcrumbList

Breadcrumb markup should follow the trail a reader actually sees. Google recommends a typical user path to the page rather than a copy of the URL folders, so a post filed under one category gets that category in the trail even if the URL has no folder at all.

Product, Offer and Reviews

Product markup belongs on pages that sell or describe one product. Google treats product snippets and merchant listings as separate experiences, and a Merchant Center feed can carry the same data for stores.

Review markup is where publishers get into trouble. Google’s review snippet guidelines set these limits:

• A review has to be about one specific item, not a category or a list.
• A business can’t mark up reviews about itself on its own pages with LocalBusiness or Organization markup, and that includes embedded third-party review widgets.
• Ratings collected on other sites can’t be pulled into your own markup.
• Fake reviews and undisclosed incentivized reviews break a guideline Google added in July 2026.

A “best X” roundup with AggregateRating stars on the whole post is the usual mistake. I explained the compliant setup in how to add star ratings to WordPress posts.

FAQPage and HowTo

Both of these had their rich results taken away in stages:

• August 2023: Google limited FAQ rich results to well-known government and health sites.
• September 14, 2023: Google removed the HowTo documentation because the rich result was no longer shown.
• May 7, 2026: FAQ rich results stopped appearing in Google Search for every site.
• June 15, 2026: Google removed the FAQ structured data documentation.

A visible FAQ section is still useful for readers, and this site keeps them. The FAQPage JSON-LD behind it earns nothing in Google Search now, and Google has never listed it as an AI Overview requirement. The full history is in my guide to FAQ schema in WordPress.

I’d also turn off any plugin setting that adds FAQPage markup to every post by default. The same goes for other types a plugin offers just because it can. Google’s guidelines say the main structured data type should reflect the main focus of the page, and a recipe type on a woodworking post helps nobody.

A Clean JSON-LD Graph for WordPress

The cleanest setup is one connected graph per page, with each entity defined once and referenced by @id. These are the rules I’d follow:

1. One Organization or Person node for the site owner, with the same @id on every page
2. One Person node per real author, matching the visible byline
3. One main entity per URL, usually BlogPosting or Product
4. A BreadcrumbList only where breadcrumbs are visible
5. References by @id instead of repeated copies of the same entity
6. Nothing in the JSON-LD that a reader can’t see on the page, apart from admin details such as a logo URL

This example shows the shape for a blog post. It is a sketch to learn from, not markup to paste as is:

{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Person",
      "@id": "https://example.com/#person",
      "name": "Jane Doe",
      "url": "https://example.com/about/",
      "sameAs": ["https://www.linkedin.com/in/janedoe"]
    },
    {
      "@type": "WebSite",
      "@id": "https://example.com/#website",
      "url": "https://example.com/",
      "name": "Example Blog",
      "publisher": { "@id": "https://example.com/#person" }
    },
    {
      "@type": "WebPage",
      "@id": "https://example.com/schema-guide/#webpage",
      "url": "https://example.com/schema-guide/",
      "isPartOf": { "@id": "https://example.com/#website" },
      "breadcrumb": { "@id": "https://example.com/schema-guide/#breadcrumb" }
    },
    {
      "@type": "BlogPosting",
      "@id": "https://example.com/schema-guide/#article",
      "headline": "How to Use Schema Markup",
      "datePublished": "2026-11-09T09:00:00+05:30",
      "dateModified": "2026-11-09T09:00:00+05:30",
      "image": ["https://example.com/images/schema-guide-16x9.jpg"],
      "author": { "@id": "https://example.com/#person" },
      "publisher": { "@id": "https://example.com/#person" },
      "mainEntityOfPage": { "@id": "https://example.com/schema-guide/#webpage" }
    },
    {
      "@type": "BreadcrumbList",
      "@id": "https://example.com/schema-guide/#breadcrumb",
      "itemListElement": [
        { "@type": "ListItem", "position": 1, "name": "Blog", "item": "https://example.com/blog/" },
        { "@type": "ListItem", "position": 2, "name": "How to Use Schema Markup" }
      ]
    }
  ]
}

Every value in it has to match the page: the headline, the byline, the dates and the breadcrumb trail.

What I Found on This Site

I checked the JSON-LD on my own entity SEO audit post. This site runs Rank Math, the WordPress SEO plugin that handles titles, sitemaps and schema here, and its Pro version prints one graph with 6 nodes:

• Person
• WebSite
• ImageObject
• BreadcrumbList
• WebPage
• NewsArticle

The author and the publisher both point to the same #person @id, which is exactly what a personal site wants. 2 things weren’t perfect, and both are common:

• The headline doesn’t match the H1. The schema headline comes from the SEO title, “Entity SEO Audit: 5 Checks to Reduce Brand Confusion”, while the page shows “Entity SEO Audit: Build a Brand Google Can Verify”. Both describe the page, but the visible title is the safer source.
• The type is NewsArticle on a blog post. Google’s Article documentation covers both types, so this costs nothing in Search today. BlogPosting is still the more accurate label.

Neither of these will make or break AI Overview visibility. This is the kind of drift that builds up on any site that has run an SEO plugin for years, and it’s cheap to fix.

Duplicate and Conflicting Schema

Most broken graphs on WordPress come from more than one plugin writing schema. These are the usual collisions:

• The theme, the SEO plugin and a schema plugin each print an Organization with a different name or logo.
• An author box plugin adds a Person with a different URL than the SEO plugin’s author.
• WooCommerce, a review plugin and the SEO plugin each print Product markup with different prices or stock.
• An old FAQ block still adds FAQPage to every post.

This is the order I’d follow to clean it up:

1. An inventory of every application/ld+json block, in the page source and in the rendered page
2. One primary schema source, usually the SEO plugin
3. Overlapping schema modules switched off everywhere else
4. Names, URLs, logos and dates lined up with what the page shows
5. A fresh test of a post, a page, a category archive and a product

If you’re still choosing the plugin that owns your schema, my Rank Math vs Yoast comparison covers how each one handles it.

WordPress Rollout Checklist

Before changing any markup, I’d take a baseline so there’s something honest to compare against later:

• a verified Search Console property
• the Search generative AI setting confirmed as Include, unless excluding is a deliberate choice
• indexable templates for posts, pages and products
• an export of the Generative AI performance report, if the property has enough impressions to show it
• the top 20 URLs by organic clicks, and by generative AI impressions where available

Then I’d work through the changes in this order:

Order	Change	Effort	Why it sits here
1	Crawl and index blockers on important pages	Medium	Eligibility comes before everything else
2	One Organization or Person node with logo and sameAs	Small	Home or about page only
3	Author Person nodes that match visible bylines and bios	Medium	Follows Google’s author guidance
4	Article fields synced to the visible title, dates and images	Small	Fixes headline and date drift
5	Visible breadcrumbs plus BreadcrumbList	Small	Only if the site uses breadcrumbs
6	Automatic FAQ and HowTo markup turned off	Small	Visible FAQs can stay
7	Product and Offer accuracy	Medium to large	Stores only
8	Review markup audit	Medium	The biggest policy risk
9	Content structure on the top pages	Large	Editorial work, not schema
10	Measurement setup	Medium	Covered in the next sections

Validation Workflow

Each tool checks a different thing, so I’d use them in this order:

1. The Schema Markup Validator checks the vocabulary and the graph. It doesn’t check anything Google-specific.
2. The Rich Results Test checks eligibility for Google’s supported rich results. Critical errors come first.
3. URL Inspection in Search Console shows what Googlebot fetched and whether the page is indexed. It’s also where you request indexing after a fix.
4. The enhancement reports in Search Console track valid and invalid items after a deploy, for the features Google still shows.

None of these predicts AI Overview selection, so a clean pass tells you the markup is correct and nothing more.

Content Structure That Helps Readers

Google’s guidance for generative AI features points back to the basics:

• original, people-first content
• clear headings and sections
• important content in text that Googlebot can read
• internal links and a good page experience

It also warns against scaled content made only to catch fan-out queries, and its spam policies now apply to generative AI responses too.

These habits help a reader first, and they also make a page easier to quote:

• an answer in the first lines of a section when the heading is a question
• passages that make sense on their own, with the subject named instead of “it” or “this”
• real HTML structure, with headings in order, lists and tables for comparisons
• original evidence such as screenshots, measurements and dated observations
• a visible FAQ for genuine follow-up questions
• an updated date only when the content changed

What I wouldn’t claim is that any of this has a fixed size or count. Google says there’s no need to chunk content for AI, and nobody outside Google knows an ideal passage length. I went deeper on formatting in how to format blog posts for AI search and in my AI search content optimization checklist.

How to Measure AI Overview Visibility

Measurement is where most reports go wrong, because 5 different things get called “AI Overview traffic.” I’d keep them apart:

What you track	What it means	Where the evidence comes from
Feature presence	An AI Overview or AI Mode answer appeared for the query	Manual checks of a fixed query set
Supporting-link impression	A link to your page was shown in the feature	Generative AI performance report
Mention in the answer	The generated text used your brand or content	Manual checks only, since Search Console has no citations metric
Click	Someone clicked your link from the feature	Performance report, inside the Web search type
Visit and conversion	What happened on the site afterward	GA4, server logs and conversion tags

What Search Console Shows

Search Console now counts AI features in 2 places, and they answer different questions.

The Performance report. Impressions, clicks and position from AI Overviews and AI Mode are counted inside the normal Web search type, mixed with every other result. Google’s counting rules for these features:

• An AI Overview takes one position, and every link in it gets that same position.
• A click on an external link in an AI Overview or AI Mode counts as a click.
• A follow-up question in AI Mode counts as a new query.

The Generative AI performance report. Google announced it on June 3, 2026 and rolled it out to all sites worldwide on August 31, 2026. It shows impressions only, with no clicks or CTR. The details:

• dimensions for pages, countries, dates and devices
• search type filters for “Web: text-based” and “Web: multimodal”
• a separate report for Discover
• no data when the site doesn’t have enough generative AI impressions, or when the site is excluded through the Search generative AI setting

It is an impressions report for supporting links. I’d never present it as a citations ledger, and I’d treat third-party AI Overview trackers as estimates from their own samples, not as Google’s numbers.

A Controlled Query Panel

Search Console can’t tell you whether an AI Overview appeared and linked to someone else, so I’d add a small manual panel. This is the setup I’d use:

• 30 to 50 queries mapped to target URLs, mixing informational, comparison, how-to and brand queries
• one fixed country, with mobile and desktop recorded separately
• a logged-out browser profile to keep personalization low
• for each check: the date and time, whether an AI Overview appeared, whether your URL was a supporting link, which competitors were linked and a screenshot
• the same person following the same steps, every week or every 2 weeks

From that, 2 rates are honest to report: how often the feature appears for your panel, and how often your URL is a supporting link when it does.

Results move for reasons that have nothing to do with your markup:

• location and language
• leftover personalization
• Google’s own experiments and interface tests
• small changes in query wording
• page updates during the test
• seasonal demand
• a panel too small to show a real change

GA4 and Server Logs

GA4 has no clean “AI Overview” channel. Clicks from AI features usually arrive with a Google referrer, mixed with other organic clicks. My suggestion is to join GA4 landing-page data with the pages that gained generative AI impressions in Search Console, for the same dates. The attribution limit should sit right next to those numbers.

Server logs show Googlebot crawling, which helps with crawl hygiene. They show nothing about citations.

Testing a Schema Change

If you want to see what a markup cleanup does, I’d set it up as a small test rather than a sitewide change:

• a group of 10 to 20 URLs that get the schema cleanup and nothing else
• a control group of similar URLs with no changes
• content rewrites kept out of the schema group, or run as a group of their own
• rich result validity, search appearance and indexing issues as the main measures
• generative AI impressions on those URLs as a secondary observation

Whatever moves, I’d report it as an association. A schema fix and a jump in AI impressions in the same month don’t prove that one caused the other.

A 30, 60 and 90 Day Plan

Window	Work	What to compare
Day 0	Baselines from the Performance report, the Generative AI report, rich result status and the first panel run	Nothing yet
Days 1 to 30	Crawl and index fixes, consistent Organization, Person and Article data, conflicting schema removed, sample URLs validated and recrawled	Rich result errors before and after
Days 31 to 60	Breadcrumb, product and review cleanup, content structure on the top panel URLs, the second panel run	Invalid item counts and panel rates
Days 61 to 90	Generative AI impression trends read with their caveats, next content priorities decided	Impressions by page, panel rates and conversions

At the end, I’d write the result up in a fixed format so nobody overclaims:

• changes shipped
• URLs in the test and control groups
• rich result validity before and after
• generative AI impressions before and after, for the property and the top URLs
• query panel presence and supporting-link rates
• anything else that changed in the same period
• the claims the report doesn’t make

Troubleshooting

Most problems trace back to one layer. This is where I’d look first:

Symptom	Likely cause	What I’d check first
No supporting links anywhere on the site	Eligibility, the AI setting or low demand	The Search generative AI setting, indexed pages and whether target queries show AI Overviews at all
The Generative AI report is missing	Too few impressions, or the site is excluded	The setting, then the query panel while volume builds
Rich Results Test fails	Markup errors	Critical errors and missing required properties, then the plugin printing them
The validator passes but no rich result shows	The type isn’t a Google feature	Whether Google still supports the type, such as FAQ or HowTo
2 different Organization names in the source	A plugin conflict	Every ld+json block on the page
Review stars on blog posts	Review markup on a list or category	itemReviewed and the review guidelines
A snippet still shows after adding nosnippet	No recrawl yet, or the tag isn’t in the fetched HTML	The crawled HTML in URL Inspection
Google-Extended is blocked but the site still appears in AI Overviews	Expected behavior	The Search generative AI setting, if exclusion is the goal

FAQs on Schema and AI Overviews

Final Remarks

If you’re still unsure whether another week of schema work will move your AI visibility, I’d look at the order of the work instead. Eligibility comes first, then a clean graph that matches the page, then honest measurement. Schema sits in the middle of that list, and describing the page accurately is the job I’d ask of it.

If your site has years of plugin schema stacked on top of each other, my schema and structured data service covers the audit and the cleanup.

Disclaimer: This site is reader-supported. If you buy through some links, I may earn a small commission at no extra cost to you. I only recommend tools I trust and would use myself. Your support helps keep gauravtiwari.org free and focused on real-world advice. Thanks. - Gaurav Tiwari