Microsoft has just unveiled NLWeb, an open-source initiative that could change how we interact with websites forever — by making them conversational.
At its Build 2025 event, Microsoft introduced NLWeb, short for “Natural Language Web,” a protocol that allows any website to provide a natural-language chat interface. In other words, NLWeb gives websites the power to talk back when you type or speak a question.
This move signals a major shift in how users could experience the web in the future — through chat, not clicks.
What Is NLWeb and Why Does It Matter?
Traditionally, websites have relied on menus, forms, and keyword search bars to help users find information. But NLWeb aims to change that by enabling websites to process plain language questions and respond with relevant answers.
Imagine asking your favorite e-commerce site:
“I need a warm jacket for Quebec’s winter.”
Or a recipe site:
“Show me spicy appetizers suitable for Diwali.”
Instead of manually applying filters, clicking through categories, or endlessly scrolling, the site simply gives you an intelligent, conversational response.
According to Microsoft, NLWeb could eventually become a foundational part of the modern web — similar to how HTML revolutionized content publishing in the early days of the internet.
How Does NLWeb Work?
At a technical level, NLWeb operates using a combination of structured data (like Schema.org or RSS feeds), vector databases (such as Qdrant or Milvus), and large language models (LLMs) like OpenAI’s GPT-4o or Anthropic’s Claude.
Here’s a simplified breakdown:
- Websites index their structured data into a vector database.
- NLWeb sets up a special /ask API endpoint that receives user queries.
- The system performs retrieval-augmented generation (RAG) to pull the most relevant pieces of content.
- It then uses an LLM to generate a conversational response.
- The response is returned in a standardized format that can be displayed via a chat interface.
This means websites can retain full control over their data while offering natural-language capabilities — without relying entirely on external search engines.
Real-World Examples: NLWeb in Action
During its Build 2025 keynote, Microsoft demonstrated several live use cases:
- TripAdvisor used NLWeb to allow users to type complex travel-related questions, such as finding Father’s Day dinner spots in Seattle with specific preferences — all via chat.
- Serious Eats showed how users could request spicy and crunchy appetizers, later refining the search to a specific holiday like Diwali — getting curated recipes instantly.
- Retail sites were able to suggest winter jackets suitable for specific weather conditions in Quebec based on conversational queries.
These examples hint at the potential for NLWeb to revolutionize retail, travel, media, and beyond — industries where users often have nuanced, subjective needs that don’t fit neatly into filter categories.
Why It Matters for the Future of the Web
NLWeb arrives at a time when conversational AI is exploding across every part of the digital ecosystem. Tools like ChatGPT, Gemini, and Claude have shown that users increasingly prefer interacting with content through chat and voice.
Microsoft’s vision for NLWeb is to provide a common, open standard that lets any site — from a personal blog to a major publisher — participate in this conversational future.
One key benefit is that NLWeb is designed to be agent-friendly. As AI agents become more popular (think smart home devices or digital assistants), NLWeb provides a structured way for these agents to retrieve reliable information directly from websites — instead of relying on opaque scraping or outdated search indexes.
Expert Insights and Industry Impact
R.V. Guha, one of the key architects behind web standards like Schema.org and RSS, is leading Microsoft’s NLWeb effort. He argues that just as HTML shaped the web’s early growth, NLWeb could define the “agentic web” — where AI agents and users can interact naturally with site content.
Early adopters like TripAdvisor, Shopify, and Hearst are already exploring how NLWeb can improve user engagement and potentially drive new monetization opportunities.
Experts also point out that NLWeb offers cost advantages compared to traditional AI search platforms because it leverages existing structured data and is more lightweight than building massive proprietary indexes.
Challenges and Considerations
Of course, NLWeb isn’t without its challenges:
- Adoption: Developers and publishers will need to implement new protocols and chat interfaces to fully leverage it.
- Discoverability: Users need visual cues (like chat widgets) so they know they can converse with the site.
- Data Responsibility: Websites must ensure they handle user queries and personal data responsibly to comply with privacy laws like GDPR.
Additionally, there is the question of whether NLWeb can remain an open standard or whether it will eventually be dominated by major platforms or browser vendors — a fate that has happened with previous promising web protocols.
What’s Next?
For now, NLWeb is in its early days. The source code is available on GitHub under an MIT license, and Microsoft is encouraging developers to start experimenting.
The goal is to build a vibrant community around the protocol — and to see if it can gain traction before proprietary conversational systems take over.