Your Content Knowledge Graph Should Power More Than Search
AI is changing how customers discover your brand.
People ask ChatGPT, Copilot, Gemini, and other AI assistants what your company does, how your products compare to competitors, and whether you’re the right solution for them. Those natural language conversations increasingly influence buying decisions before someone ever visits your website.
Great content still matters, but content alone isn’t enough. AI systems need structured, trustworthy information they can understand and reference with confidence.
For years, Schema App has helped enterprise organizations build governed Content Knowledge Graphs. Using schema.org, we transform website content into a machine-readable source of truth that search engines can understand.
As conversational AI became another way customers interact with brands, we saw an opportunity. The same Content Knowledge Graph that improves search visibility can also power conversational AI experiences.
Our new product, Knowledge Assistant, brings that vision to life.
Powered by NLWeb, Knowledge Assistant allows users to ask natural-language questions directly on your site and receive accurate, cited answers directly from the organization’s Content Knowledge Graph.
This article explains why we chose NLWeb, how Knowledge Assistant works, and what we’ve learned by running it ourselves.
Why We Built Knowledge Assistant on NLWeb
When Microsoft introduced NLWeb, we immediately saw how well it aligned with our approach.
Our customers had already invested in building governed Content Knowledge Graphs. What they needed was a practical way to make that knowledge available through natural language without creating another disconnected knowledge base.
NLWeb provided exactly that. It gave us a standard way to retrieve trusted information, generate responses, and expose the same knowledge across conversational experiences, websites, and emerging AI agents.
Knowledge Assistant became the product that brings those capabilities together in a customer-ready experience.
NLWeb serves as the underlying technology and infrastructure: retrieval, synthesis, citation, and deployment surfaces. While Knowledge Assistant is the customer-facing product: the conversational UI in Schema App, plus paths to hosted experiences and agent endpoints. See it in action below:
How Knowledge Assistant Works
Every question follows the same four-step process.
1. Understand the question
In today’s search experiences powered by AI, people ask questions naturally. They often describe problems, compare products, or ask for recommendations using everyday, long-form human language.
Knowledge Assistant interprets the user’s natural language and intent instead of expecting them to know your navigation, product names, or documentation structure.
2. Retrieve the right information
Semantic retrieval identifies the entities and content most relevant to the question using the customer’s Content Knowledge Graph.
Because the graph defines not only individual entities but also the relationships between them, Knowledge Assistant understands how products, services, documentation, FAQs, people, and locations connect. That allows it to answer questions that span multiple pages or domains without losing business context, supporting enterprise scale and complexity.
Semantic retrieval helps Knowledge Assistant find the right information quickly. The Content Knowledge Graph provides the business context that keeps those answers accurate.
For organizations with multiple websites, this becomes especially valuable. A single conversation can seamlessly combine information from marketing sites, support documentation, or other digital properties because they’re all connected through the same governed knowledge layer.
3. Generate a grounded response
An LLM assembles an answer using the retrieved information. If sufficient information is available, the response includes citations to the original content.
If the answer doesn’t exist, Knowledge Assistant says so. For enterprise organizations, acknowledging uncertainty is far more valuable than generating an unsupported answer. It protects brand trust while giving teams a clear signal that additional content or structured data is needed.
For enterprise organizations, those unanswered questions often become the most valuable output because they reveal opportunities to strengthen content, expand structured data, and improve the Content Knowledge Graph over time.
4. Present information in useful ways
Structured data improves the experience beyond the answer itself.
Because Knowledge Assistant understands the structure of your content, it can present information differently depending on what it retrieves, as rich results.
- FAQ content can appear as expandable question-and-answer sections.
- Contact information becomes an actionable routing option.
- Questions that span multiple products or services can synthesize information from across your website while preserving citations back to the original sources.
Instead of treating every response as another paragraph of text, Knowledge Assistant presents information in ways that reflect how your business is organized and how users naturally navigate information.
Why This Matters for Enterprise Teams
Enterprise organizations aren’t looking for another chatbot. They’re looking for a reliable way to extend the value of the semantic data layer they’ve already invested in creating.
Knowledge Assistant helps teams:
- Answer customer questions using brand-aligned, governed information
- Reuse existing structured data investments instead of creating another disparate AI knowledge base
- Identify content gaps through a database of real customer questions
- Support websites, support centers, and future AI experiences from the same semantic foundation
The same structured data that supports search and rich results can now support on-site conversations and future AI agents. Teams govern their Content Knowledge Graph and make it available across multiple experiences, rather than maintaining separate systems for each new interface.
What We’ve Learned Running Knowledge Assistant Ourselves
Using Knowledge Assistant across schemaapp.com and support.schemaapp.com has changed how we think about preparing content for AI.
1. Customers ask broader questions than search ever revealed
Search data often reflects keywords, but conversational interfaces reveal intent.
People ask questions like:
- Which product is right for my business?
- How do these features compare?
- Who should I contact about this?
- Where do I find setup instructions?
Many of these questions span multiple products, documentation pages, or website sections. They reinforced the importance of connecting information through entities rather than treating every page independently.
2. Every evaluation becomes a content audit
One of the greatest advantages Knowledge Assistant offers enterprise teams is its ability to identify where customers expect answers that don’t yet exist. Those unanswered questions quickly become a prioritized roadmap for improving content, expanding structured data, and strengthening the Content Knowledge Graph.
3. Transparency builds trust
Users and enterprise teams want to know where information comes from. Showing citations gives them confidence in the answer while helping content owners verify that approved information is being surfaced accurately.
4. One semantic data layer supports many experiences
The same Content Knowledge Graph now supports search, Knowledge Assistant, and emerging AI interfaces and integrations. As new interfaces emerge, organizations can extend the same governed knowledge rather than rebuilding it for each new platform.
One Knowledge Layer for Every AI Experience
Search, conversational experiences, and AI agents all depend on the same thing: trusted, machine-readable data. The interfaces will continue to evolve, but the underlying knowledge should not.
Knowledge Assistant demonstrates how the Content Knowledge Graph that many organizations already use for search can become the foundation for conversational AI. NLWeb provides the standard that makes those conversations possible while allowing the same knowledge to support future experiences as they emerge.
We think about that future in terms of three outcomes:
- Authority. AI answers from your governed, brand-aligned Content Knowledge Graph rather than piecing together information from across the web or making assumptions.
- Accuracy. Every answer is grounded in your approved content, supported by citations, and transparent about what it doesn’t know.
- Trust. Marketing, product, legal, and customer support teams can have greater confidence that AI is representing the business using approved information.
Every major shift in search has rewarded organizations that invested in better information instead of short-term tactics. We believe AI and agents will follow the same pattern.
That’s why we’re continuing to invest in Knowledge Assistant. Not simply to build another AI experience, but to help enterprise organizations put their Content Knowledge Graph to work wherever customers choose to ask questions.
Book a demo of Knowledge Assistant today.
Already a customer and want to learn more? Talk to your Customer Success Manager about access to Knowledge Assistant.
Mark van Berkel is the Chief Technology Officer and Co-founder of Schema App. A veteran in semantic technologies, Mark has a Master of Engineering – Industrial Information Engineering from the University of Toronto, where he helped build a semantic technology application for SAP Research Labs. Today, he dedicates his time to developing products and solutions that help enterprise teams structure and connect their data so it is accurately understood by search engines and AI, improving visibility and enabling more effective AI-driven outcomes.