Inferensys

Glossary

ClaimReview

A Schema.org type used to markup a fact-check of a specific statement, allowing publishers to indicate the claim being reviewed and its veracity rating, which search engines use to surface fact-check labels in results.
Developer reviewing semantic search engine results on laptop, relevance scores visible, technical search demo.
FACT-CHECKING STRUCTURED DATA

What is ClaimReview?

ClaimReview is a Schema.org type that enables publishers to markup a fact-check of a specific statement, allowing search engines to surface a veracity rating label directly in results.

ClaimReview is a CreativeWork subtype in the Schema.org vocabulary designed to annotate a published analysis of a specific claim. It formally connects the claim text, the author of the fact-check, the date of review, and a machine-readable reviewRating (e.g., True, False, Misleading) into a single structured data object.

Search engines like Google and Bing ingest ClaimReview markup to power fact-check labels in search results and news surfaces. For a ClaimReview to be valid, it must reference a specific, checkable statement made by a distinct entity, not a general topic, and the publisher must be a verified, authoritative fact-checking organization.

STRUCTURED DATA FACT-CHECKING

Core Properties of ClaimReview Markup

The essential Schema.org properties that define a fact-check, linking a specific claim to its veracity rating and enabling search engines to surface fact-check labels in results.

01

claimReviewed

The exact textual statement being fact-checked. This property holds the claim itself, not a summary or paraphrase. For accurate entity extraction, the claim should be a short, self-contained assertion.

  • Best practice: Quote the claim verbatim as it appeared in the original source.
  • Example: "The Earth is flat and stationary."
  • Critical for: Enabling search engines to match the fact-check to the original claim across the web.
02

reviewRating

The veracity assessment of the claim, expressed as a Rating type. This is the core output of the fact-check, communicating the truthfulness of the claimReviewed.

  • Standardized values: Use a defined scale like True, False, Mostly True, Mixture, or custom ratings.
  • Best practice: Always include both ratingValue (the text label) and alternateName (a short human-readable version).
  • Example: "ratingValue": "False" with "alternateName": "Pants on Fire".
03

author

The Organization or Person that conducted the fact-check and published the assessment. This property establishes the credibility and authority of the review.

  • Must be an entity: Link to an Organization schema with a name, url, and ideally a sameAs reference to a Wikidata or Wikipedia entry.
  • Transparency signal: Search engines use this to determine if the fact-checker is a verified, trusted source.
  • Example: "author": {"@type": "Organization", "name": "Snopes.com", "url": "https://www.snopes.com"}.
04

itemReviewed

The CreativeWork (article, social media post, video) that contains the claim being fact-checked. This property contextualizes the claim by linking it to its source material.

  • Nested entity: Use a CreativeWork type with properties like author, datePublished, and appearance (the URL where the claim appeared).
  • Distinction: itemReviewed is the container of the claim; claimReviewed is the specific assertion within it.
  • Example: A viral Facebook post containing a false statistic would be the itemReviewed.
05

url

The canonical URL of the full fact-check article. This is the permanent link where users can read the complete analysis, methodology, and evidence behind the rating.

  • Required for rich results: Google requires this to display the fact-check label in search results.
  • Best practice: Ensure the URL is a dedicated, non-paginated page with the full fact-check content.
  • Critical for: Driving traffic to the authoritative source and enabling users to verify the assessment independently.
06

datePublished

The ISO 8601 date when the fact-check was first published. This temporal marker is critical for assessing the freshness and relevance of the review.

  • Format: "2024-11-15T08:00:00+00:00"
  • Why it matters: A fact-check of a rapidly evolving news story may become outdated. Search engines prioritize recent reviews for current claims.
  • Best practice: Update this date only if the fact-check is substantively revised, not for minor typo fixes.
CLAIMREVIEW EXPLAINED

Frequently Asked Questions

Clear, technical answers to the most common questions about implementing and understanding the ClaimReview schema for fact-checking markup.

ClaimReview is a Schema.org type specifically designed to markup a fact-check of a statement. It works by wrapping a CreativeWork (the fact-check article) with structured data that identifies the exact claim being reviewed, the entity that made the claim, and a definitive veracity rating. Search engines like Google and Bing parse this markup to surface a Fact Check label directly in search results, displaying the claim, the rating, and the publisher. The mechanism relies on the claimReviewed property to hold the verbatim text of the statement, while reviewRating uses a numeric scale (typically 1-5) and a textual label (e.g., "False," "Mostly True") to communicate the conclusion. This structured approach allows platforms to algorithmically aggregate fact-checks across the web without manual curation.

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.