Beyond 2D: How AI Breaks Traditional Web Discovery

Overview

This research examines how AI agents navigate websites, where they succeed, and where they fail. Using 3,348 navigation tasks across 269 websites, we identify patterns that challenge conventional assumptions about AI-optimised content.

Key Findings

• Search is bimodal: Search-first navigation either works immediately (90% success) or fails badly (27% success when navigation is required)
• The Two-Click Rule: 91% of successful navigation completes within two clicks
• Position trumps relevance: AI agents click links in early DOM positions 80% of the time, regardless of semantic match
• Failure means loops: 95% of failed traces involve the agent revisiting pages it's already seen
• Industry variation: Success rates range from 68% (Pharma/Healthcare) to 84% (SaaS)

Implications

The web has entered a third dimension. Content must exist, rank well, AND be navigable by AI agents. Traditional SEO gets you in the game; AI accessibility determines whether you win.

Methodology

Each trace represents a single AI agent attempting to complete a specific task on a website. We tested three navigation approaches: homepage-first, search-first, and hybrid. The dataset includes 494,197 link evaluations across 77 task categories and 7 industries.