Schema.org for AI Surfacing: Which Markups Matter Now

Schema.org for Artificial Intelligence Surfacing: Which Markups Matter Now

As of September 5, 2026, structured data still helps search engines understand pages, authors, organizations, and facts. However, it is not a direct ranking switch for artificial intelligence answers.

Google states that pages do not need special Schema.org markup to appear in AI Overviews or AI Mode. A page must mainly be crawlable, indexed, eligible for a search snippet, and supported by useful content. Google also says structured data should match the visible page content. (developers.google.com)

The best current strategy is therefore:

1. Use structured data to describe the page accurately.
2. Match the markup to the page's true purpose.
3. Build clear relationships between articles, authors, organizations, and topics.
4. Write direct, complete answers in visible HTML.
5. Measure artificial intelligence citations separately from traditional rich results.

Executive verdict

The broad research finding is important: adding generic structured data alone has not produced a consistent increase in artificial intelligence citations. A controlled Ahrefs study tracked 1,885 pages that added JavaScript Object Notation for Linked Data and compared them with 4,000 control pages. It found no meaningful improvement in Google AI Mode or ChatGPT citations. Google AI Overview citations declined slightly, but the researchers warned that the change was small and could not clearly be blamed on the markup. (ahrefs.com)

A separate 2026 preprint found that generic types such as Article, Organization, BreadcrumbList, and WebPage did not independently predict artificial intelligence citations after controlling for search rank and domain authority. Its strongest finding was that pages with concrete, attribute-rich data, such as prices, ratings, and specifications, performed better than pages with only generic page labels. That finding focused mainly on product and review pages, so it should not be treated as proof that any one of the types in this article creates a citation advantage. (aixiv.science)

What structured data can and cannot do

Structured data is a machine-readable description of a page. It can tell a search engine:

• What type of page it is
• Who wrote it
• Which organization published it
• What question it answers
• What date it was published or updated
• Which person, company, term, or dataset the page describes

Google says structured data can help its systems understand page content and make pages eligible for richer search features. It also says that Google Search may use other Schema.org properties for understanding, even when those properties do not trigger a visible search result. (developers.google.com)

Structured data does not guarantee:

• A higher organic ranking
• An artificial intelligence citation
• A rich result
• A knowledge panel
• Inclusion in an artificial intelligence answer
• Use of the exact text in the markup

Bing gives similar guidance. Its current webmaster guidance says structured data may support clearer grounding, but it does not guarantee visibility or citation traffic. Bing also advises publishers to make facts and definitions explicit in the visible page content. (bing.com)

The main research limitation

Artificial intelligence answer panels usually show the source page, not the Schema.org type that may have been present on that page. Google does not publish a report saying, for example, that a page was cited because it used Article instead of WebPage.

This creates three different questions:

1. Was the page cited?
2. Did the page contain structured data?
3. Did the structured data cause the citation?

Most studies can answer only the first two. They cannot prove the third.

That is why a page with FAQPage markup may appear frequently in artificial intelligence answers without the markup being the reason. The page may have strong content, a high search ranking, many links, or a well-known brand.

Audit by schema type

1. Article

What it does

Article describes an article, news story, blog post, or similar editorial page. Google supports Article, NewsArticle, and BlogPosting as article types. Google does not list required properties for article markup, but it recommends adding the properties that apply to the page. (developers.google.com)

Properties that matter most

Use these when they are visible and accurate:

• headline
• author
• author.name
• author.url or author.sameAs
• datePublished
• dateModified
• image
• publisher
• mainEntityOfPage
• about
• inLanguage

Google recommends using a real Person or Organization for the author. It also recommends keeping dates in structured data consistent with the visible publication and update dates. (developers.google.com)

Artificial intelligence effect

Evidence level: indirect.

Article helps establish page type, authorship, and freshness. These are useful signals for search systems, especially on facts pages and editorial content. However, current evidence does not show that adding Article alone increases artificial intelligence citations.

Article checklist

• The page is genuinely an article.
• The headline matches the visible title.
• Every visible author is included.
• Each author has a separate Person or Organization object.
• Author names contain names only, not job titles or publisher names.
• The author links to a real profile or author page.
• Publication and update dates are visible on the page.
• Dates use the correct time zone when time is included.
• The image represents the article.
• The publisher is identified consistently across the site.
• The article is not marked up as a different primary type, such as HowTo, unless the page truly serves both purposes.

2. WebPage

What it does

WebPage is a general page type. Schema.org states that every web page is implicitly treated as a WebPage, but an explicit declaration can help when the page includes page-level properties or relationships. (schema.org)

Useful properties include:

• url
• name
• description
• inLanguage
• dateModified
• breadcrumb
• mainEntity
• about
• isPartOf
• primaryImageOfPage

Artificial intelligence effect

Evidence level: low and indirect.

WebPage is best used as the outer page layer in a connected graph. It can connect the page to its main article, definition, dataset, person, or organization.

It should not be treated as a special artificial intelligence optimization type. A page that contains only a generic WebPage object usually provides less useful information than a page that clearly identifies its main entity.

WebPage checklist

• Use one stable @id for the page.
• Use the canonical URL as the page URL.
• Identify the page's true mainEntity.
• Link the main entity back to the page with mainEntityOfPage.
• Add inLanguage when known.
• Keep the page name and description aligned with visible content.
• Do not use WebPage to hide the fact that the page is really an article, profile, dataset, or question page.

3. QAPage

What it does

QAPage is for a page focused on one question and its answers. Google says it uses Question structured data from pages marked as QAPage, and there should be only one QAPage and one main Question on the page. (developers.google.com)

Required properties

For current Google question-and-answer eligibility:

• QAPage.mainEntity
• A nested Question
• Question.answerCount
• Either acceptedAnswer or suggestedAnswer
• Answer.text

A question with no answers is not eligible for the rich result.

Important content rule

Do not use QAPage for:

• A normal frequently asked questions page
• A blog post that answers a question
• A how-to article
• A product page containing many questions
• An editorial answer written only by the site owner

Google says users must be able to submit answers for a normal QAPage. Valid examples include a forum question or a support page where users can provide answers. (developers.google.com)

Artificial intelligence effect

Evidence level: medium semantic fit, no proven causal lift.

A real question-and-answer page is naturally easy for a retrieval system to understand. However, no strong public study proves that QAPage markup itself increases artificial intelligence citations.

QAPage checklist

• The page focuses on one question.
• Users can submit answers, unless the page qualifies for a special education question-and-answer experience.
• The full question is visible.
• The full answer text is visible.
• answerCount matches the actual number of answers.
• Accepted and suggested answers are labeled correctly.
• Comments are marked as comments, not answers.
• The page is not simply an editorial frequently asked questions page.
• The page does not contain multiple unrelated questions.

QAPage example

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "QAPage",
  "@id": "https://www.example.com/questions/ounces-in-a-pound#qapage",
  "mainEntity": {
    "@type": "Question",
    "@id": "https://www.example.com/questions/ounces-in-a-pound#question",
    "name": "How many ounces are in one pound?",
    "text": "How many ounces are in one pound?",
    "answerCount": 1,
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "There are 16 ounces in one pound.",
      "author": {
        "@type": "Person",
        "name": "Jordan Lee",
        "url": "https://www.example.com/users/jordan-lee"
      },
      "datePublished": "2026-08-20T10:00:00-04:00"
    }
  }
}
</script>

Use this pattern only when the page truly supports a question-and-answer interaction.

4. HowTo

What it does

HowTo describes step-by-step instructions. Google once supported How-to rich results, but it deprecated that search feature in September 2023. Google said How-to results would no longer appear on desktop and had already been removed from mobile search. (developers.google.com)

Artificial intelligence effect

Evidence level: low for Google.

The visible steps may still help users and retrieval systems. A clear tutorial with headings, numbered steps, tools, time, and warnings is easier to read and quote. But current evidence does not show that HowTo markup creates a special advantage in Google AI Overviews or AI Mode.

Recommendation

Use HowTo only when:

• The page genuinely teaches a task.
• The steps are visible in the page content.
• Another search engine, platform, or internal system benefits from the markup.
• Your team can maintain it without creating conflicting data.

For Google Search, prioritize strong HTML headings, numbered lists, clear instructions, and useful images or video.

Tutorial checklist

• The page teaches a real task.
• The result of the task is clear.
• Each step is visible and complete.
• Step names match the visible headings.
• Tools and supplies are real and visible.
• Time estimates are accurate.
• Safety warnings are included where needed.
• The first section gives a short answer or outcome.
• The page does not rely on markup to provide the instructions.

5. ClaimReview

What it does

ClaimReview was designed for fact-checking content. Google phased out Claim Review support in Search as part of its 2025 effort to simplify search results. The type was removed from Search Console reporting and the Rich Results Test. (developers.google.com)

Artificial intelligence effect

Evidence level: no current Google advantage.

A high-quality fact-check can still be cited because it clearly states:

• The claim
• The rating
• The evidence
• The date
• The fact-checking organization
• The reasoning behind the conclusion

Those benefits come mainly from the content itself, not from the retired Google search feature.

Recommendation

For a facts page:

1. Use Article or NewsArticle when the page is editorial.
2. Clearly state the claim in visible text.
3. Cite primary evidence.
4. Identify the author and reviewing organization.
5. Add publication and review dates.
6. Use ClaimReview only if another platform or data system specifically requires it.

Do not add ClaimReview only because you expect Google artificial intelligence answers to prefer it.

6. FAQPage

What it does

FAQPage describes a page containing questions and official answers. Google stopped showing the FAQ rich result in Search starting May 7, 2026, and removed the related documentation in June 2026. (developers.google.com)

Artificial intelligence effect

Evidence level: weak and mixed.

A 90-day vendor study added FAQPage markup to 120 pages. It found no reliable improvement in ChatGPT, Gemini, or Google AI Overview citations. Perplexity showed a small increase, but the study itself said the result was platform-specific and did not prove causation. (authorityradar.com)

Another study of 615 already-cited pages found that FAQ markup appeared more often on heavily cited pages. That relationship disappeared after controlling for repeated pages from the same publishers. The researchers concluded that the evidence did not establish an effect from the markup itself. (getintel.ai)

Recommendation

Use frequently asked questions when they improve the page for readers. Do not add large blocks of generic questions just to target artificial intelligence answers.

If you keep FAQPage markup for another search engine or content system:

• Make every question visible.
• Make every answer complete.
• Keep the markup identical to the page.
• Do not repeat the same question in several schema blocks.
• Do not expect a Google FAQ rich result.

FAQPage example for non-Google consumers

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "@id": "https://www.example.com/support/cloud-backups#faq",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How often should cloud backups run?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "For important business files, schedule automatic backups at least once each day. Use more frequent backups when the files change often."
      }
    },
    {
      "@type": "Question",
      "name": "How long should backups be kept?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Keep backups long enough to recover from accidental deletion, system failures, and security incidents. Many organizations keep daily backups for 30 to 90 days."
      }
    }
  ]
}
</script>

This is a semantic description, not a promise of a Google search feature.

7. Organization

What it does

Organization helps Google understand and disambiguate a company, nonprofit, publisher, school, or other organization. Google says organization markup can influence visual elements such as the logo shown in Search and some knowledge panel information. There are no required properties in Google's current organization guide. (developers.google.com)

Recommended properties

Use the properties that are true and visible:

• name
• alternateName
• url
• logo
• sameAs
• description
• telephone
• email
• address
• identifier
• foundingDate
• parentOrganization

Artificial intelligence effect

Evidence level: indirect but useful.

Organization can connect:

• The publisher to an article
• The company to its products or services
• The brand to its official profiles
• The organization to a known web identity

This is useful for entity disambiguation. It does not prove that an artificial intelligence system will cite the page.

Organization checklist

• Place the full organization object on the home page or organization page.
• Use a stable @id, such as https://www.example.com/#organization.
• Use the exact public organization name.
• Link to real official profiles with sameAs.
• Use the correct organization subtype when appropriate.
• Use a real logo that represents the organization.
• Keep contact information current.
• Reference the organization from articles instead of recreating conflicting versions on every page.

8. Person

What it does

Person identifies a person who writes, reviews, owns, manages, or appears on a page. It is usually most useful when connected to:

• Article.author
• QAPage question or answer author
• ProfilePage.mainEntity
• Organization.employee
• Review.author

Google's profile guidance says a profile page must focus on one person or organization. The ProfilePage object requires a mainEntity, and that entity must be a Person or Organization. The person or organization must have a name, or an alternateName when no name is available. (developers.google.com)

Recommended properties

• name
• url
• sameAs
• image
• description
• jobTitle
• worksFor
• knowsAbout
• affiliation
• identifier

Artificial intelligence effect

Evidence level: indirect.

Person markup can help connect an author's name to:

• A biography
• A job or role
• An organization
• Published articles
• External profiles
• Areas of expertise

Use it to make identity clear, not to claim expertise that the page does not support.

Person checklist

• Use Person only for a real person.
• Use Organization for a company or publication.
• Link the person to a visible author page.
• Use sameAs only for accurate, official profiles.
• Keep job titles and credentials current.
• Add all visible authors, not just the lead author.
• Use the same person @id across articles and profile pages.

Required property matrix

Google's general guidance favors complete and accurate data over large amounts of incomplete markup. It also warns that structured data must represent visible content and that correct markup still does not guarantee a rich result. (developers.google.com)

Use-case implementation checklists

Facts pages

Best combination:

• WebPage
• Article or NewsArticle
• Person
• Organization
• Optional ClaimReview only for another supported consumer

Checklist:

• State the main fact near the top of the page.
• Name the source of the fact.
• Link to primary evidence.
• Include the date of publication and last review.
• Identify the author and reviewer.
• Separate facts from opinion.
• Use Article when the page is editorial.
• Do not use ClaimReview as a current Google Search tactic.

Definition pages

Best combination:

• WebPage
• DefinedTerm
• Optional Article if the page is a long editorial explanation
• Organization or Person when an expert or publisher is responsible

DefinedTerm is intended for a word, phrase, code, or concept with a formal definition. Its main properties include name, description, termCode, inDefinedTermSet, and sameAs. (schema.org)

Checklist:

• Give the definition in the first paragraph.
• Use one clear term as the main entity.
• Add alternate names only when they are real.
• Link to a reliable external definition when appropriate.
• Explain the term in plain language.
• Use examples and boundaries.
• Avoid marking a list of unrelated terms as one DefinedTerm.

Tutorials

Best combination:

• WebPage
• HowTo only when another consumer needs it
• Article when the tutorial is also an editorial article
• Person and Organization for authorship

Checklist:

• State the result before the steps.
• Use numbered visible headings.
• Keep each step focused on one action.
• Include tools, supplies, time, and warnings where needed.
• Add images or video when they help.
• Do not hide the steps only in JSON-LD.
• Do not expect How-to rich results in Google Search.

Data catalogs

Best combination:

• WebPage
• DataCatalog
• Dataset
• DataDownload
• Organization

Schema.org defines Dataset as a body of structured information and supports relationships such as includedInDataCatalog and distribution. (schema.org)

Google clarified in late 2025 that Dataset structured data is used by Dataset Search and is not a general Google Search result feature. It should therefore be treated as a data discovery and interoperability layer, not an artificial intelligence citation shortcut. (developers.google.com)

Checklist:

• Give each dataset a stable identifier.
• State the subject and scope.
• Include the publisher or creator.
• Add the date range covered by the data.
• State geographic coverage when relevant.
• Describe licenses and access conditions.
• Add each downloadable file as a DataDownload.
• Include file format and download URL.
• Keep catalog metadata synchronized with the actual files.
• Document update frequency and last update date.

JSON-LD example: facts page

This example connects the page, article, author, publisher, and topic. Replace every value with information that appears on the real page.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "WebPage",
      "@id": "https://www.example.com/facts/renewable-energy-storage#webpage",
      "url": "https://www.example.com/facts/renewable-energy-storage",
      "name": "Renewable Energy Storage: Key Facts",
      "inLanguage": "en-US",
      "dateModified": "2026-08-20T09:00:00-04:00",
      "mainEntity": {
        "@id": "https://www.example.com/facts/renewable-energy-storage#article"
      },
      "about": {
        "@id": "https://www.example.com/facts/renewable-energy-storage#term"
      }
    },
    {
      "@type": "Article",
      "@id": "https://www.example.com/facts/renewable-energy-storage#article",
      "mainEntityOfPage": {
        "@id": "https://www.example.com/facts/renewable-energy-storage#webpage"
      },
      "headline": "Renewable Energy Storage: Key Facts",
      "description": "A fact-checked overview of how renewable energy storage works.",
      "image": [
        "https://www.example.com/images/energy-storage-16x9.jpg",
        "https://www.example.com/images/energy-storage-1x1.jpg"
      ],
      "author": {
        "@id": "https://www.example.com/authors/maya-chen#person"
      },
      "publisher": {
        "@id": "https://www.example.com/#organization"
      },
      "datePublished": "2026-07-10T08:00:00-04:00",
      "dateModified": "2026-08-20T09:00:00-04:00",
      "about": {
        "@id": "https://www.example.com/facts/renewable-energy-storage#term"
      }
    },
    {
      "@type": "Person",
      "@id": "https://www.example.com/authors/maya-chen#person",
      "name": "Maya Chen",
      "url": "https://www.example.com/authors/maya-chen",
      "jobTitle": "Energy Research Editor",
      "worksFor": {
        "@id": "https://www.example.com/#organization"
      },
      "sameAs": [
        "https://www.linkedin.com/in/maya-chen"
      ]
    },
    {
      "@type": "Organization",
      "@id": "https://www.example.com/#organization",
      "name": "Example Research Group",
      "url": "https://www.example.com/",
      "logo": "https://www.example.com/images/logo.png",
      "sameAs": [
        "https://www.linkedin.com/company/example-research-group"
      ]
    },
    {
      "@type": "DefinedTerm",
      "@id": "https://www.example.com/facts/renewable-energy-storage#term",
      "name": "Renewable energy storage",
      "description": "Methods used to store energy generated from renewable sources for later use.",
      "url": "https://www.example.com/facts/renewable-energy-storage"
    }
  ]
}
</script>

JSON-LD example: definition page

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "WebPage",
      "@id": "https://www.example.com/glossary/retrieval-augmented-generation#webpage",
      "url": "https://www.example.com/glossary/retrieval-augmented-generation",
      "name": "Retrieval-Augmented Generation: Definition",
      "inLanguage": "en-US",
      "mainEntity": {
        "@id": "https://www.example.com/glossary/retrieval-augmented-generation#term"
      }
    },
    {
      "@type": "DefinedTerm",
      "@id": "https://www.example.com/glossary/retrieval-augmented-generation#term",
      "name": "Retrieval-augmented generation",
      "alternateName": "RAG",
      "description": "A method in which an artificial intelligence system retrieves relevant information before generating an answer.",
      "url": "https://www.example.com/glossary/retrieval-augmented-generation",
      "inDefinedTermSet": {
        "@type": "DefinedTermSet",
        "name": "Example Research Group Glossary",
        "url": "https://www.example.com/glossary"
      },
      "mainEntityOfPage": {
        "@id": "https://www.example.com/glossary/retrieval-augmented-generation#webpage"
      }
    }
  ]
}
</script>

The definition must also appear as normal page text. Do not place the definition only in the structured data.

JSON-LD example: tutorial

Because Google's How-to rich result is deprecated, treat this as optional markup for other systems. The visible page should still contain the full instructions.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "WebPage",
      "@id": "https://www.example.com/tutorials/create-a-data-dictionary#webpage",
      "url": "https://www.example.com/tutorials/create-a-data-dictionary",
      "name": "How to Create a Data Dictionary",
      "mainEntity": {
        "@id": "https://www.example.com/tutorials/create-a-data-dictionary#howto"
      }
    },
    {
      "@type": "HowTo",
      "@id": "https://www.example.com/tutorials/create-a-data-dictionary#howto",
      "name": "How to Create a Data Dictionary",
      "description": "A step-by-step guide to documenting fields in a data system.",
      "totalTime": "PT45M",
      "tool": [
        {
          "@type": "HowToTool",
          "name": "Spreadsheet software"
        }
      ],
      "step": [
        {
          "@type": "HowToStep",
          "name": "List the fields",
          "text": "Create a list of every field in the data system.",
          "url": "https://www.example.com/tutorials/create-a-data-dictionary#step-1"
        },
        {
          "@type": "HowToStep",
          "name": "Describe each field",
          "text": "Record the field name, meaning, data type, allowed values, and owner.",
          "url": "https://www.example.com/tutorials/create-a-data-dictionary#step-2"
        },
        {
          "@type": "HowToStep",
          "name": "Review the dictionary",
          "text": "Ask a subject matter expert to check the definitions and examples.",
          "url": "https://www.example.com/tutorials/create-a-data-dictionary#step-3"
        }
      ]
    }
  ]
}
</script>

JSON-LD example: data catalog

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "WebPage",
      "@id": "https://data.example.com/catalog#webpage",
      "url": "https://data.example.com/catalog",
      "name": "Example Public Data Catalog",
      "mainEntity": {
        "@id": "https://data.example.com/catalog#catalog"
      }
    },
    {
      "@type": "DataCatalog",
      "@id": "https://data.example.com/catalog#catalog",
      "name": "Example Public Data Catalog",
      "description": "A catalog of public datasets about transportation and local services.",
      "url": "https://data.example.com/catalog",
      "publisher": {
        "@id": "https://data.example.com/#organization"
      },
      "dataset": [
        {
          "@id": "https://data.example.com/datasets/transit-ridership#dataset"
        }
      ]
    },
    {
      "@type": "Dataset",
      "@id": "https://data.example.com/datasets/transit-ridership#dataset",
      "name": "Monthly Transit Ridership",
      "description": "Monthly passenger counts for public transportation services.",
      "url": "https://data.example.com/datasets/transit-ridership",
      "identifier": "transit-ridership-monthly",
      "creator": {
        "@id": "https://data.example.com/#organization"
      },
      "publisher": {
        "@id": "https://data.example.com/#organization"
      },
      "license": "https://creativecommons.org/licenses/by/4.0/",
      "temporalCoverage": "2020/2026",
      "spatialCoverage": {
        "@type": "Place",
        "name": "Example County"
      },
      "includedInDataCatalog": {
        "@id": "https://data.example.com/catalog#catalog"
      },
      "distribution": [
        {
          "@type": "DataDownload",
          "contentUrl": "https://data.example.com/files/transit-ridership.csv",
          "encodingFormat": "text/csv",
          "name": "Comma-separated values download"
        }
      ]
    },
    {
      "@type": "Organization",
      "@id": "https://data.example.com/#organization",
      "name": "Example County Data Office",
      "url": "https://data.example.com/"
    }
  ]
}
</script>

Common implementation pitfalls

Mismatched schema

The most serious mistake is marking up content that users cannot see. Google says structured data must be a true representation of the page, and misleading or hidden content can make a page ineligible for rich results. (developers.google.com)

Common examples:

• Marking an article as a HowTo when it contains no real steps
• Marking a company as an author when the article was written by a person
• Adding FAQ answers that do not appear on the page
• Using a future publication date
• Marking a general blog post as QAPage
• Adding ClaimReview to an opinion article

Thin answers

Structured data cannot fill an empty page.

A short, vague answer inside Answer.text or acceptedAnswer does not create a strong source. The visible content should:

• Answer the question directly
• Explain important limits and exceptions
• Name sources
• Include dates, examples, or measurements where useful
• Stand on its own when copied out of context

Google's artificial intelligence guidance says there is no ideal page length and no need to break content into tiny pieces for artificial intelligence systems. The better goal is useful, complete, people-first content. (developers.google.com)

Duplicate entities

Avoid publishing several conflicting versions of the same organization, author, or page.

Weak implementation:

• One Organization object with one name on the home page
• A second object with a different name on every article
• A third object with no @id on the author page

Better implementation:

• Give the organization one stable @id
• Give each author one stable @id
• Reference those objects from articles, profiles, and question pages
• Keep the name, logo, URL, and external identity links consistent

Duplicate questions

Do not repeat the same question in:

• FAQPage
• QAPage
• Article markup
• Several visible page sections
• Multiple JSON-LD blocks

Use the schema type that matches the page's main purpose. A single clear answer is better than several overlapping markup blocks.

Incorrect dates

Google uses several sources to estimate publication and update dates. It recommends that visible dates and structured dates agree, and it warns against using future dates or dates related to events discussed in the article rather than dates related to the page itself. (developers.google.com)

Overusing sameAs

A sameAs link should identify the same real-world person or organization. Do not link to:

• An unrelated social profile
• A search results page
• A generic directory listing
• A page with a different spelling or identity
• A profile that the organization does not control

JavaScript-only markup

Google can process structured data added to the rendered page, but a JavaScript-only implementation can be harder for other crawlers and auditing tools to detect. A server-rendered JSON-LD block is usually easier to test and maintain. (developers.google.com)

A practical testing plan

To measure whether markup has an incremental effect, use a controlled test instead of relying on a few manual searches.

Before the change

Record:

• Target queries
• Current organic ranking
• Whether an artificial intelligence answer appears
• Which pages are cited
• Citation position when available
• Search traffic
• Conversions
• Current structured data
• Content changes made during the test period

During the test

• Add one major markup change at a time.
• Keep content, internal links, titles, and backlinks stable.
• Use similar control pages that do not receive the change.
• Record the exact publication date of the change.
• Wait long enough for crawling and reprocessing.

Ahrefs used matched controls and a before-and-after difference-in-differences method. Its approach is a useful model for organizations that want to test structured data instead of assuming that a correlation proves causation. (ahrefs.com)

After the change

Track:

• Google Search Console artificial intelligence performance data
• Google AI Overview citations
• Google AI Mode citations
• Bing Webmaster Tools artificial intelligence citations
• ChatGPT, Gemini, or Perplexity citations when relevant
• Organic rankings
• Search clicks
• Assisted conversions

Google reports artificial intelligence search traffic through Search Console performance reporting. Bing's artificial intelligence performance reporting shows cited pages and grounding queries, but it does not show why a page was selected or how important it was within an answer. (developers.google.com)

Recommended implementation order

For most publishers, the best order is:

1. Fix visible content first.
2. Make crawling and indexing reliable.
3. Implement Article for real editorial pages.
4. Connect authors with Person and profile pages.
5. Connect publishers with Organization.
6. Use WebPage as a clean page-level graph layer.
7. Use QAPage only for genuine community questions.
8. Use DefinedTerm for glossary and definition pages.
9. Use Dataset and DataCatalog for data resources.
10. Treat FAQPage, HowTo, and ClaimReview as secondary or non-Google markup because their Google search features have been removed or deprecated.

Conclusion

The strongest current lesson is simple: Schema.org markup helps machines understand content, but it is not a guaranteed path into artificial intelligence answers.

The most durable implementation is not a large collection of schema types. It is a small, accurate entity graph:

• Article describes the editorial page.
• Person identifies the author.
• Organization identifies the publisher.
• WebPage connects the page to its main entity.
• QAPage describes a genuine user question and its answers.
• DefinedTerm clarifies a definition.
• Dataset and DataCatalog describe structured data resources.

Use structured data where it adds clear meaning. Do not use it to disguise thin content, duplicate visible text, or imitate a search feature that Google no longer supports. For artificial intelligence surfacing, the highest-value work remains clear answers, strong evidence, accurate entities, current information, and content that can stand on its own.