The web as we know it is changing, but the direction is uncertain

Recent announcements around Model Context Protocol (MCP) and initiatives like Microsoft's NL Web underscore a significant shift in how we might soon interact with the web and the internet as we know it. MCP, a new standard from Anthropic (the company behind Claude), essentially acts as a structured API replacement, enabling AI agents to interact seamlessly with websites and databases without traditional user interfaces that are used by humans.

This technology prompts a provocative thought: could websites as we know them, with user interfaces built for humans, become redundant? Rather than browsing traditional web pages, agents could directly retrieve and process data from databases, requiring only minimal human intervention for tasks like authentication or payments. While exciting, this scenario raises essential questions about the future business models and implications for the open web.

Microsoft’s NL Web initiative offers a contrasting vision, where AI capabilities are integrated directly into individual websites. Rather than relying solely on centralised chatbots such as ChatGPT, users would interact directly with websites enhanced with their own AI functionalities. This decentralised approach is appealing, particularly for those who, like myself, value an open and interconnected internet.

However, tech giants such as Google and Facebook seem largely indifferent to preserving traditional web content. The focus of platforms like Facebook has shifted dramatically, with news content comprising only a minor fraction of it’s overall user engagement, thereby reducing incentives to maintain those avenues. Similarly, Google appears increasingly focused on viewing the web primarily as an interconnected series of databases, with less emphasis on preserving websites in their current forms.

This shift potentially disrupts current web monetisation strategies, particularly advertising-based models (goodbye SEO?). The implications could be profound for many businesses, potentially forcing a dramatic restructuring of how online content and services are funded. Publishers, retailers, and traditional online businesses might soon find themselves needing to transition rapidly into API-driven models to stay relevant.

The debate also highlights contrasting perspectives on web centralisation. While MCP's vision might initially centralise access to information through AI agent registries, there is also an emerging argument for greater decentralisation, exemplified by the ever growing federated social networks and distributed content platforms. The hope is that these new technologies could reverse the web’s historical trend towards centralisation, empowering smaller, independent players rather than reinforcing the dominance of a few large corporations.

Yet significant hurdles remain. The infrastructure required to support comprehensive AI agent interactions-registries, discoverability mechanisms, and user-friendly interfaces-is still nascent. While powerful AI demonstrations generate enthusiasm, practical applications remain limited and primarily enterprise-focused for now.

My optimistic outcome could be a rejuvenated, decentralised web, compelling websites to evolve into deliberate destinations offering unique experiences, rather than merely transient endpoints from a Google search. Conversely, the pessimistic scenario sees websites losing their distinct identities, reduced to interchangeable databases for AI queries.

For advocates of the open web, including myself, the coming years represent a critical juncture. How these technologies are adopted and regulated will shape not only the digital landscape but also broader economic and social interactions online. As we navigate these transformative times, maintaining a balance between innovation, decentralisation, and preserving the best aspects of the traditional web will be paramount.