Web iv is the intelligent and agentic web

We are at the start of something big. Web iv, or Web 4, is the intelligent and agentic web. Software increasingly interprets what we want, plans how to achieve it and acts on our behalf.

This is a thesis I’ve been building through my work on agents, discovery and data exchange. The commercial question is who captures the value.

Here is how I define the generations. They overlap; these are descriptions of changing behaviour, not official software releases.

• Web 1.0: static. Following Tim Berners-Lee’s invention at CERN in 1989, the 1990s web grew around linked pages. Publishers published; most visitors read.
• Web 2.0: connected and social. In the 2000s, participation became central. People created, shared and collaborated through blogs, wikis, social networks and web applications. Platforms organised the activity.
• Web3: blockchain and decentralisation. The 2010s brought a vision of distributed ownership, using blockchains, tokens and smart contracts to reduce dependence on central platforms.
• Web iv: intelligent and agentic. In the 2020s, AI systems increasingly interpret information, plan and use tools. People delegate tasks across websites and services. Early computer-use systems demonstrate this direction, even while their reliability remains limited.

Web3 did not live up to its hype. My judgement is that it failed to deliver the promised decentralised everyday web. It largely served a technical cohort, with too much attention on who could make the most money from crypto and NFTs. Building a web that was useful and accessible to everyone felt secondary.

For me, the major milestones are 1, 2 and now 4: publishing, participation and delegation.

The terminology is contested. The European Commission uses Web 3.0 for the semantic web and Web 4.0 for the convergence of AI, connected devices and virtual worlds. I consider that virtual-world framing irrelevant to the transition described here. Agents changing how work gets done is enough to mark a new era.

The pattern I see with each major iteration is a worse bargain for independent webmasters. More commercial value consolidates inside the largest platforms. The evidence is becoming difficult to ignore:

• Content consumed, little traffic returned. Across sites behind Cloudflare in June 2025, Google made about 14 crawls per referral; OpenAI made 1,700 and Anthropic 73,000. These are platform-level ratios, and missing native-app referrals can inflate them. Even with that limitation, the imbalance is striking. Cloudflare analysis.
• Fewer clicks to websites. In Pew’s study of 900 US adults’ March 2025 browsing, traditional results received clicks on 8% of Google visits with an AI summary, versus 15% without. Only 1% of visits with a summary produced a click on its citations. This is an observed association, not a causal estimate. Pew research.
• Audiences and data accumulate. Meta reported 3.58 billion daily users across its apps in December 2025, up 7% in a year. The FTC found that major social and video platforms’ business models encouraged mass data collection for monetisation, particularly targeted advertising. Meta results, FTC findings.

We should push back. A fair and equitable Web iv needs open standards, meaningful choice over how content and data are used, and value returned through attribution, referrals, reciprocal access or payment. That is the ambition behind my work on Odexa.

We need to rally for a web where innovation can thrive beyond the largest technology platforms, and creators share in the value they generate. A more intelligent web should reward the people who make it worth using.