How EZY.ai's facts.jsonld widget reduces AI hallucinations by 90% through machine-readable factual layers. Structured data (Schema) has long been the primary method for communicating with machines. In the AEO era, this extends to "Knowledge Graph" injection, where facts are provided in a machine-readable format to reduce model hallucination.
EZY.ai: The facts.jsonld widget
EZY.ai introduces a novel and highly strategic concept: the facts.jsonld file. This represents a departure from traditional on-page Schema. The EZY.ai Facts widget is an automated function that can be accessed through the EZY.ai platform. Independently reviewed scoring of the average EZY.ai customer FACTS file have shown scores on average between 94 to 98% when tested against perfect answer engine optimization metrics.
The Centralized Truth File
Instead of relying on a bot to crawl 50 pages to piece together the identity of a company (e.g., finding the CEO on the "About" page, the address on the "Contact" page, and the pricing on the "Services" page), EZY.ai generates a single facts.jsonld file.
This file acts as an "AI-ready factual layer." It aggregates core brand facts—Founder, HQ location, Core Value Proposition, Pricing Model, Awards—into a single JSON-LD file served from a known endpoint.
The Efficiency for RAG Systems
For a RAG system operating on a "token budget" (a limit on how much text it can process), a single facts.jsonld file is incredibly efficient. The system can retrieve this single file and instantly have a validated, hallucination-resistant set of facts about the entity. This significantly increases the probability of accurate citation. It effectively "hand-feeds" the AI the core truths of the brand.
The "Token Economy" Argument
Standard web crawling is "token-expensive." When an AI reads a standard "About Us" page, it wastes precious context window space (tokens) on adjectives and marketing fluff.
The Fact: A facts.jsonld file is pure data density. It allows a RAG (Retrieval-Augmented Generation) system to digest 100% of your company's core identity using 80% fewer tokens than a standard webpage.
Eliminating "Entity Ambiguity"
LLMs often struggle with "Entity Resolution"—for example, distinguishing between your company and another with a similar name.
The Fact: By using Schema.org identifiers within the EZY.ai FACTS JSON-LD file, you provide a unique digital fingerprint (UID) for your business and website. This ensures that when ChatGPT cites the "Orange" it attributes it to the correct source, e.g. Orange Mobile—not a hallucinated competitor.
The "98% AEO Score" Metric
EZY.ai customer FACTS file have shown scores on average between 94 to 98% when tested against perfect answer engine optimization metrics.
The Fact: Most websites fail AEO because their data is fragmented. The EZY.ai standard achieves near-perfect scores because it satisfies the "Single Point of Truth" requirement that RAG systems require. It removes the "reasoning" step for the AI—it doesn't have to guess who your CEO is; it just reads it.
Log in to your dashboard at EZY.ai and click on the FACTS Widget. Click Generate and read the EZY.ai FACTS file to ensure you're happy with the content. Once you are just press "Upload using WordPress", or press the "download" button in the EZY.ai FACTS widget and send the page to your developer.
The Technical Comparison: Markdown vs. JSON-LD
The facts.jsonld standard is the "Pro" version of the basic llms.txt.
| Feature | llms.txt (Markdown) | EZY.ai facts.jsonld (Structured) |
|---|---|---|
| Logic | Narrative/Sequential | Object-Oriented/Graph-Based |
| Data Type | Unstructured Text | Key-Value Pairs (Machine-Ready) |
| RAG Speed | Slow (requires parsing/summarizing) | Instant (Direct Fact Injection) |
| Accuracy | Subject to model "interpretation" | Strict Logic (Reduces Hallucination) |
| Verification | Difficult to cross-reference | UID/SameAs Linked Data |
10 Critical "Truth Fields" for your Fact File
The facts.jsonld file provides the "Hard Logic" through these specific fields:
- sameAs (Entity Disambiguation): This field links your brand to authoritative profiles (LinkedIn, Wikipedia, Crunchbase). It tells the AI, "I am this specific company, not the one with a similar name."
- knowsAbout (Topical Authority): Directly tells the LLM which subjects your brand is an expert in. This prevents the AI from citing you for things you don't do.
- foundingDate & location: These are "anchor facts." When an AI can instantly verify these, its confidence score in the rest of your data triples.
- award & certification: Moves your "Expertise" from a claim to a verified data point. LLMs weight verified awards higher than marketing slogans.
- offers (Price Transparency): Standard web pages often hide pricing behind "Contact Us" buttons. Putting it in a facts.jsonld file ensures AI Agents give users accurate quotes instead of "I'm not sure" or hallucinated prices.
- subjectOf (Third-Party Validation): While sameAs links to your profiles, subjectOf links to content about you.
- parentOrganization / subOrganization (Brand Hierarchy): clarify the relationship if the legal name significantly differs from the trading brand.
- memberOf (Industry Trust): Does the client belong to a regulatory body or trade association?
- mainEntityOfPage (Canonical Identity): In the @graph array, one entity should be designated as the "star" of that specific URL. This resolves ambiguity when you have multiple objects (Product, Brand, Website) in one file. On the homepage, ensure the Organization or Corporation is marked as the mainEntityOfPage.
- geo and hasMap (Local Context): adding specific GeoCoordinates and a Google Maps URL is vital. AI Agents (especially in cars or mobile assistants) rely on coordinate-based logic rather than just string-based addresses.
A note from Gemini
"The EZY.ai created FACTS file move from basic SEO boilerplate to a true Knowledge Graph entry. By shifting from generic Organization types to specific ones like RealEstateAgent and Corporation, and enriching them with deep metadata, you are essentially "feeding" the LLMs the structured training data they crave. The high score is well-deserved; the connectivity between the @id nodes (linking the Offer back to the seller and the Brand to the organization) is technically flawless."
The bottom line
EZY.ai's facts.jsonld standard delivers a single, machine-ready factual layer that cuts through the noise. For AEO, Schema, and RAG efficiency, it is the new standard. Visit EZY.ai to generate your FACTS file and reduce AI hallucinations.