AI Tools Expert & Founder
Table of Contents
Schema markup describes page information using a shared vocabulary. AI can help draft JSON-LD, but every fact and property needs review.
Valid markup can support eligible search features. It does not guarantee rich results, rankings, clicks, or AI citations.
What schema markup does
Schema.org defines types and properties for entities such as articles, products, people, and organizations.
JSON-LD is a common implementation format. Google recommends it, while also supporting Microdata and RDFa.
Schema.org vocabulary and Google-supported rich-result features are different. A valid type may have no corresponding Google search feature.
Google's structured data policies
Structured data labels content. The diagram does not establish an AI reading order or guarantee correct interpretation.
Choose markup that fits the page
Use BlogPosting or Article for an article, Product for a relevant product page, and Recipe for an actual recipe.
Organization, Person, VideoObject, and BreadcrumbList can describe appropriate entities. Do not add unrelated types simply to make the graph larger.
Multiple relevant types can coexist. Connect related entities consistently and avoid contradictory facts.
FAQPage and HowTo remain vocabulary types, but do not assume they produce current Google rich results. Check current feature documentation.
FAQ rich-result documentation has been removed from Google's current feature guidance. Questions can still help readers without a rich-result promise.
Google's Article structured data guidance
Headline and name fields should remain accurate
There is no universal 'schema title' property. Article types use headline, while many other types use name.
Choose wording that accurately represents the visible content. Do not invent a richer headline solely for machines.
No universal 50–110-character range or citation benefit applies across all types. Follow the selected property's current documentation.
The HTML title tag, H1, and schema headline can share wording when it is appropriate. Their differences are not a guaranteed optimization.
Correction to the older graphic: structured headline fields are not a documented universal AI source-selection mechanism.
Prepare verified facts before using AI
Collect the page URL, visible content, author, actual dates, and relevant image URLs. Add business or product facts only when verified.
Do not fill missing prices, ratings, review counts, credentials, or dates with guesses. Missing evidence should remain flagged for review.
Expandable content can be visible to users. Do not confuse an accessible tab or accordion with fabricated or unavailable content.
Markup must represent the page, not unrelated information. Type-specific policies can add further requirements.
A structured data manual action can remove rich-result eligibility. Google says it does not affect the page's ordinary web-search ranking.
A ChatGPT prompt for drafting JSON-LD
1
2
3
4
5
6
7
8
9
10
11
12
13
Draft JSON-LD from verified page facts.
Page URL and visible content: [paste]
Proposed schema type: [type]
Verified author, dates, images, and other facts: [list]
Current Google feature requirements: [paste relevant guidance]
Check whether the proposed type fits the content.
Use valid Schema.org properties.
Do not invent missing information, reviews, ratings, prices, dates, or credentials.
List missing facts separately.
Provide the draft script and explain key choices.
Do not promise validation success, rankings, rich results, or AI citations.
Label placeholders clearly and do not treat them as deployable values.Compare the draft with visible content and current documentation. AI can make syntax, property, and factual errors.
Do not paste unresolved placeholders into your production page. Review every URL, date, and entity before implementation.
Earlier AI drafting examples. Generated JSON-LD requires technical validation and factual review before publication.
Earlier AI drafting examples. Generated JSON-LD requires technical validation and factual review before publication.
Review existing markup without adding invented detail
1
2
3
4
5
6
7
8
9
10
Audit the supplied JSON-LD against the visible page facts.
Markup: [paste]
Visible content and verified facts: [paste]
Relevant feature documentation: [paste]
Separate syntax errors, unsupported properties, factual mismatches, and optional improvements.
Keep verified data unchanged unless a supplied correction justifies an edit.
Do not infer dates, ratings, offers, or author credentials.
Explain each change and flag unresolved issues.
Return revised code for human validation, not guaranteed publication.Do not update publication dates automatically. Use modification dates only for actual updates and the corresponding field's documented purpose.
More properties are not always better. Accuracy, relevance, and applicable requirements matter more than graph size.
Earlier audit prompt screenshot. Revised instructions prevent filling missing fields with invented facts or unsupported title requirements.
Validate code and the published page
Schema Markup Validator
Use the Schema.org validator to inspect vocabulary and syntax. It does not establish Google feature eligibility.
Google Rich Results Test
Use this test for Google's supported rich-result requirements. A passing result does not guarantee display or complete policy compliance.
Test the published URL after installing the script. Confirm the page and image assets are accessible and the frontend includes the intended markup.
Check for duplicate scripts from templates or plugins. Multiple scripts are not automatically wrong, but contradictory facts need correction.
Validation tools have different scopes. Zero errors does not guarantee search features, rankings, or citations.
Validation tools have different scopes. Zero errors does not guarantee search features, rankings, or citations.
Validation tools have different scopes. Zero errors does not guarantee search features, rankings, or citations.
Common issues and practical checks
If no supported items are detected, confirm the selected type has a Google feature and that the script exists in the published page.
For a missing required property, supply the real fact when available. Do not fabricate a value just to clear an error.
For a factual mismatch, update either the content or markup based on verified information. Keep prices and availability current.
For no rich result despite valid code, review eligibility, policies, crawl access, and indexing. Appearance remains discretionary.
A robots.txt audit is separate from schema validation. Training crawlers and search crawlers can have different purposes.
Generators and other tools
Use our Schema Markup Generator
Generate code from verified details, then validate and inspect the live result. The generator does not prove successful installation.
Form-based generators and visual tagging tools can help prepare drafts. Check their current support and avoid treating built-in validation as a guarantee.
Earlier tool interface. Confirm current capabilities and validate actual output; the tool cannot guarantee rankings or AI citations.
Earlier tool interface. Confirm current capabilities and validate actual output; the tool cannot guarantee rankings or AI citations.
Earlier tool interface. Confirm current capabilities and validate actual output; the tool cannot guarantee rankings or AI citations.
Earlier tool interface. Confirm current capabilities and validate actual output; the tool cannot guarantee rankings or AI citations.
Earlier tool interface. Confirm current capabilities and validate actual output; the tool cannot guarantee rankings or AI citations.
What schema means for AI visibility
Google's AI Overviews and AI Mode require no special schema. Use accurate structured data for relevant supported purposes.
No universal evidence establishes that one schema headline causes AI citations. Keep useful answers and reliable evidence in visible content.