A new survey of AI practitioners finds that the web is fracturing into two distinct layers — one for humans, one for machines — and the transition is moving faster than the industry anticipated.
The next major shift in artificial intelligence may not be the model. It may be the architecture of the web itself.
Bright Data released research on Sunday showing that as AI agents become a primary interface for search, retrieval, and action, the internet is beginning to divide into two layers: one built for human browsing, and another engineered for agentic access. Based on a February 2026 survey of 500 AI practitioners, the report found that 71 percent of organizations are already deploying AI agents for web search, and 87 percent agreed that a “two-tier internet” is taking shape. Half of the respondents expect this bifurcation to solidify within two years.
The transition is placing new strain on the systems that power AI. As organizations consume more data and grow dependent on real-time information for production AI, the bottleneck is shifting away from model performance. The more pressing constraint, practitioners say, is the infrastructure required to retrieve, validate, and govern live external data at scale. Survey respondents reported using 132 percent more data to train models over the past 12 months. Among the top obstacles: 57 percent cited data quality and validation as their foremost challenge in enabling real-time reasoning.
Access to public web data is tightening at the same time. Eighty-eight percent of respondents said public web data is becoming more restricted through gatekeeping measures. Seventy-three percent anticipate increased regulatory legislation, and 58 percent expect website blocking to rise in the near future.
Also Read: Amazon’s Layoffs Aren’t a Warning. They’re a Preview.
The findings point to a new competitive reality for AI developers: succeeding in the agentic era will require investment in web data infrastructure, not just stronger models.
“We are watching the web evolve from a place humans browse into an environment agents must navigate on their own,” said Or Lenchner, chief executive of Bright Data. “That changes the infrastructure requirements completely. In the agentic era, the advantage goes to companies that can access real-time public web data reliably, validate it rigorously, and do it compliantly at scale.”
The full report, “Data for the Agentic Web,” is available on Bright Data’s website.
