- Mar 12
- 6 min read
Originally released as a Microsoft-led open project, NLWeb allows creators to make their websites queryable by both humans and AI agents. Instead of clicking through links and navigating menus, users can ask questions directly to your site and receive structured, context-aware answers derived from your own data.
What is NLWeb?
NLWeb, or natural language web, is an open-source conversational search platform designed to bridge the gap between traditional websites and the agentic web by enabling natural language interactions with websites. NLWeb turns websites into conversational interfaces. Think of it like a combined search engine and AI chat bot, trained on your website data. It can be accessed on your domain or via tools like Claude Code.
It’s based on two crucial ideas:
Search should be both conversational and agentic. It should allow humans and AI agents to ask complex questions, take actions, and navigate across an ecosystem of NLWeb-enabled sites to complete multi-step tasks.
We need to move beyond keyword-focused SERPs to interfaces designed for how LLMs actually work, feeding information to AI systems in structured, context-aware formats rather than forcing them to parse traditional search results. If we were building the ideal search engine from scratch today, we wouldn’t start with 10 blue links.
As Chelsea Carter said in her Human in the Loop interview with Wix Studio, rather than thinking about keywords and page rankings, we're focused on creating endpoints that can understand and respond to complex natural language queries. We’re building the infrastructure that enables publishers to participate in the agentic web without requiring them to become AI experts.
How does NLWeb work?
NLWeb turns the public data on your website into an AI-friendly layer that users, agents, and LLMs can search using natural language. This means that agents can tap into data from a website’s schema.org, product feeds, and RSS to more accurately answer conversational searches via the NLWeb WebApp of their site.
Instead of relying on scraping or custom APIs, NLWeb reads the data on your site and converts it into a unified grounding source that humans and AI agents can understand.
At the core of NLWeb is a semantic indexing pipeline that takes your structured content—like schema markup, RSS feeds, and product data—and generates vector embeddings that capture the meaning behind your data. This then allows NLWeb to interpret natural‑language questions with nuance: a query about “pricing” can surface information labeled as “cost,” “plans,” or “subscription tiers,” even if those exact words don’t appear on the page.
At its core, NLWeb is about standardizing how websites respond to natural-language queries. Just as HTML created a common format for rendering web pages, NLWeb defines a consistent way for agents to ask questions and receive structured answers.
How does NLWeb use structured data on your website?
NLWeb uses the structured data on your website as a source to generate better quality answers to questions from human users and AI agents.
When a user or AI agent asks a question, NLWeb retrieves the most relevant pieces of your structured data and returns them in a predictable, machine‑readable format.
Because the system is built on open standards, and can leverage MCP, any compatible agent can interact with your site without custom integration work—your content becomes instantly accessible and up‑to‑date.
By using the structured data you already maintain, NLWeb turns your website into an interoperable, AI‑native application ready for the agentic web. It gives agents a reliable way to understand and query your domain, making your content discoverable and actionable through natural‑language interaction.
The type of structured data used in NLWeb
While search marketers talk about “structured data” they’re often using it as shorthand for schema.org structured data markup. But in actuality, structured data is a term that applies to any data that has been tagged and/or organized.
On a website, this can be your schema.org structured data markup, but it can also be your RSS feed, product feed, and other data sources. More than 45 million web domains use schema.org to help search engines understand their pages. NLWeb goes a step further by using this, and other forms of structured data, to answer questions about your domain.
Why the NLWeb matters for marketers
NLWeb converts your website from a passive archive into an authoritative data source that feeds directly into the AI models and agents your customers are using. By implementing NLWeb, you ensure your site is not just a destination for browsers, but a source of truth that AI agents can interact with directly. NLWeb activation is valuable for number of reasons...
Faster route to agent readiness: Marketers no longer need to wait for months-long development cycles to create agent-ready access points on their websites Because NLWeb leverages existing data like schema.org, it provides immediate functionality without complex prompt engineering, allowing you to deploy a sophisticated AI interface in a matter of hours.
Reduced friction: Visitors get answers immediately without digging through content archives, complex navigation, or having to speak in exact keywords on native search. This is valuable for customer retention and accessibility.
Data freshness: The data you provide to LLMs is based on your content like RSS and product feeds, which automatically update with content changes. This means that NLWeb responses can tap into the most current version of your site, ensuring users receive up-to-the-minute information without manual updates.
Scalability: NLWeb's technology-agnostic design allows you to use your website's data as a unified, up-to-date source for building and deploying custom, customer-facing AI features like CustomGPTs and virtual assistants on any external AI model, including platforms like ChatGPT or Claude, without requiring a backend rebuild.
How to optimize for NLWeb
NLWeb works best when your site already uses clean structure, clear entities, and schema markup. The goal is to transform your site into a "source of truth" by reducing the effort required for an LLM to parse your information. Start by auditing your site's structure and content clarity. Pages should answer specific questions clearly and consistently.
Technical NLWeb optimizations
Implement and audit structured data: Ensure your schema markup is valid and complete for products, services, FAQs, and organizations. This helps AI agents understand the entities and relationships on your site and helps NLWeb to more accurately reflect your site.
Optimize your RSS and content feeds: Review your RSS and product feeds for content alignment and robustness. Since NLWeb uses these feeds to stay updated, ensure they contain full-text content and high-quality metadata rather than just headlines. This allows AI agents to access your latest updates in real-time without waiting for a search engine to re-crawl your site.
Select an AI-native CMS platform: Use a platform that supports technical SEO and structured data by default. Your CMS should automatically generate the JSON-LD and clean XML sitemaps that NLWeb-powered agents rely on to navigate your site’s architecture without hitting technical bottlenecks.
Content optimizations
Clean up your content: Delete or update inaccurate and out-of-date content. Vague or duplicated copy limits how well AI can form accurate answers, and consistency across pages matters more than ever.
Design content for answers, not just clicks: Use clear headings, concise explanations, and strong internal consistency to help AI understand and extract information from your pages.
Which websites can use NLWeb?
NLWeb isn't limited to one type of business. NLWeb early adopters include media, travel, food, and technology brands. Wix is a platform partner for the NLWeb and is working to leverage NLWeb across Wix, Wix Studio, and Wix Harmony websites.
So, any organization with structured content can use it to make their data more accessible to users and AI agents.
Here are some examples of how different industries can leverage NLWeb.
eCommerce sites: Customers can ask specific questions like, "show me waterproof hiking boots under $150 in size 10" and receive instant, filtered results. AI shopping assistants can also compare products across specifications, check real-time inventory, and complete purchases on behalf of users.
Publishing and media companies: Readers can query archives conversationally ("Find articles about climate policy from the past year") and allow AI research assistants to pull relevant stories, quotes, and data directly from their content library with proper attribution.
SaaS companies can turn their documentation and knowledge bases into conversational support systems, allowing both human users and AI agents to troubleshoot issues, find API references, and access setup guides without navigating through multiple help center pages.
Local businesses and service providers (restaurants, salons, medical practices) can let customers ask about availability, services, and pricing, then book appointments through natural conversation. With NLWeb, information is also accessible to AI assistants handling scheduling and recommendations.
Whether you're selling products, publishing stories, or providing services, the opportunity is the same: make your website conversational. The question isn't whether AI agents will browse the web differently than humans do—it's whether your site will be ready when they arrive.
