Dataset coverage

What this dataset includes.

This is a frozen measured citation panel from Aug 2, 2026. It is not a live view of the Web or all AI answers.

Measured prompts

The dataset contains 60,406 canonical prompts. Of those, 21,320 have at least one valid citation edge and form the denominator for the public source results.

Prompts outside that eligible group cannot be assumed to have received a successful answer with zero citations; some may lack a usable extraction.

Public evidence

Public results show approved website, topic, and safe cited-page evidence. Prompt text and unsafe URL details are withheld. A missing source in this panel is not proof that it never appears elsewhere.

What we cannot measure

This release does not establish real-user query volume, traffic, source quality, crawling, training use, causal answer contribution, prices, or licensing value. It also cannot support reliable model-by-model or time-trend claims.

Read the methodology for how observations are counted.