Learning Center/Foundations

How AI usage becomes metrics

Every step an AI takes with your content fires a usage event — and those events are where publisher metrics come from. Learn the six events, then watch a live agent fire them.

Live demo · 5 questions per session

Step 1 · Learn the events

The six usage events

One question, end to end: a reader asks an AI assistant “What are the early symptoms of Lyme disease?” Here’s each event as it happens.

10,000,000 documents
catalog scopeCashmere sees
10 million documents — about 8 billion tokens — sit on the assistant’s API key. Nothing has been used yet; this is the shelf.
24 passages
content_retrievedCashmere sees
The agent runs three searches; 24 passages (about 9,000 tokens) leave the API. Cashmere meters every one automatically.
6 passages
content_groundedApp-reported
The model decides 6 passages actually answer the question and writes from those (about 2,400 tokens). The other 18 are set aside — only the app knows which.
3 sources
content_citedApp-reported
The answer credits 3 of the 6 grounded sources as [1] [2] [3]. The other three shaped it without credit.
2 of 3 shown
content_displayedApp-reported
The sidebar surfaces only 2 of the 3 cited sources. The third is credited in the text but never given a card — so the reader can never click it.
1 click
content_engagedCashmere sees
The reader clicks one card to check the original — landing on the Cashmere-hosted source page, measured automatically.

Cashmere is helping define and shape the Content Telemetry standard with the SPUR Coalition — an open standard for measuring how AI systems use content.

Step 2 · See it live

Now connect it to how the AI operates

Watch a real agent fire these events live — and see which metric each one moves.

Live demo · 5 questions per session

A live demo against public sample collections on Cashmere. Grounded, cited, and displayed are reported by this page itself; in production they exist only if the application emits these events.

Up nextHow AI actually uses your contentAn interactive walkthrough: follow one question from your catalog to a cited answer, and see what those tokens are worth.