Ramanathan V. Guha, a technical fellow at Microsoft and a key figure behind web standards like RSS and Schema.org, recently introduced a new open web protocol called NLWeb at Microsoft Build 2025 conference. He frames NLWeb as part of the fourth major revolution in personal computing following graphical user interfaces, the internet, and mobile technology. This new revolution centers on communicating with computers and applications using natural, free-form language, not just artificial intelligence alone.
Guha is concerned about the current AI landscape, where much interaction is dominated by centralized chatbots like ChatGPT, Claude, and Bing, which often just aggregate knowledge without adding real value. His vision with NLWeb is to empower any website or app owner to easily add AI-powered conversational search features using just a few lines of code, their own data, and a chosen AI model. NLWeb is a protocol that takes natural language questions and returns structured answers, handling the technical complexity so developers can focus on their data and expertise. This approach can create smarter, more customized chatbots tailored to specific sites rather than relying on general-purpose AI.
For example, Guha demonstrated how a site like Serious Eats could use NLWeb to offer personalized recipe suggestions based on user preferences like vegetarianism or holiday-specific needs. Similarly, an outdoor retailer could provide product recommendations that consider local weather conditions. These AI-powered searches are not only more relevant but also cheaper to run than traditional web search, which requires costly crawling and indexing. NLWeb leverages existing semi-structured data formats like RSS and Schema.org combined with vector databases and affordable AI models (like GPT-4o Mini) to deliver fast, low-cost natural language search.
Microsoft is collaborating with companies such as TripAdvisor, Eventbrite, and Shopify to implement NLWeb, aiming to make conversational AI a native feature of the web rather than a service controlled by a few large AI providers. NLWeb is also connected to the Model Context Protocol (MCP), an open project by Anthropic, which helps integrate AI features into websites. Microsoft’s broader goal is to encourage widespread adoption of AI conversational tools, with the expectation that many businesses will eventually pay for Azure-based AI services. However, Guha emphasizes that NLWeb is designed as an open, technology-agnostic protocol that anyone can use and extend.
Other articles you may find interesting
Despite its promise, NLWeb faces challenges. The web has historically trended toward centralization, with dominant players like Google and large AI platforms controlling search and chatbots. Convincing major companies like Meta and Google to adopt an open protocol may be difficult, as many prefer proprietary solutions or licensing deals. Additionally, questions remain about privacy, data sharing, and how AI agents might act autonomously on users’ behalf, such as making purchases or sharing personal preferences across sites. Guha admits that the future of NLWeb depends on how the web community chooses to develop and govern it.
In essence, NLWeb aims to democratize AI-powered search and conversational interfaces, making them accessible and affordable for all websites. If successful, it could transform how people interact with the web, shifting from centralized AI chatbots to a more decentralized, open ecosystem where each site can offer intelligent, personalized experiences. This could mark a significant step in the ongoing evolution of the internet, much like how HTML and RSS once did.
NLWeb’s reliance on existing web standards and its open design may encourage innovation by developers and smaller publishers who previously lacked resources to implement advanced AI features. It also aligns with broader industry trends toward “agentic” AI intelligent agents that can perform tasks autonomously on behalf of users, potentially reshaping e-commerce, customer service, and content discovery. However, this also raises important ethical and security considerations about AI autonomy and user data control that the community will need to address as NLWeb and similar protocols evolve.
Summary
In summary, NLWeb represents a bold attempt to make AI-powered natural language interaction a fundamental, decentralized part of the web, empowering websites to offer smarter, more personalized experiences without relying solely on large centralized AI platforms. Its success will depend on adoption by the web ecosystem and how well it balances innovation with privacy and control.
