Which software pricing do AI assistants research most?

What the data shows

Four patterns stand out in the 30-day window, each computed directly from the table below rather than asserted.

1. Developer and AI infrastructure dominate. The top of the ranking is almost entirely tools that engineers integrate — Higgsfield AI, Bloomberg Terminal, Cloudflare Zero Trust, and their peers in hosting, databases, and LLM APIs. The five busiest categories (AI Video Generators, Financial Data & Terminals, Developer Tools, AI Chatbots, LLM API Providers) together account for 11.9% of all retrievals. People ask assistants about the pricing of tools they build on far more than about the pricing of everyday consumer apps.
2. Machines read these records ~127× more than humans visit them. Across every mapped product, assistants performed 73,863 user-initiated retrievals but only 583 humans then arrived on a record from an assistant — a retrieval-to-referral ratio of about 127:1. The value of being in an assistant’s answer is mostly in being the machine’s source, not in click-through traffic. That is exactly why we never blend the two numbers.
3. Referral yield varies sharply by category. Among the busiest categories, LLM API Providers converts retrievals into human visits far more often (22 AI-referred visitors on 1,269 retrievals) than AI/GPU Cloud Compute (3 on 1,039). Consumer-facing AI categories tend to send people onward; developer-infrastructure categories are mostly read by the machine and rarely clicked.

These are single-window observations, not trends. We are publishing one 30-day snapshot; we do not yet make trend claims. See the methodology for what each metric does and does not mean.