Skip to main content

AI Usage > Sessions

Written by Tom Williams

Dataset: AI Usage

Entity: AI Usage Measurement

Field ID: sessions

Type: Number

Description: The number of distinct working sessions. Not interchangeable with requests or interactions. Most providers publish only some of the three, so an empty value is missing coverage rather than a zero.

Source: App

Transformation logic: N/A

App Mapping

GitHub Copilot

session_count on command line and desktop app records; on agent app records it counts agent job starts

Anthropic Claude (coming soon)

core_metrics.num_sessions

Cursor (coming soon)

the number of distinct conversations in the record

OpenAI Platform (coming soon)

N/A

OpenAI Codex (coming soon)

threads

Reporting Use Cases

The Sessions field counts distinct working sessions - a conversation, a thread, a command line session. It is the coarsest of the three volume measures and the most stable across tool changes.

  • Habit rather than volume: Sessions per active developer per week is a good measure of whether AI use has become routine, and it is far less sensitive than requests to which surface someone happens to use.

  • Session depth: Dividing interactions by sessions shows whether people are having short exchanges or long working sessions.

  • Different meanings per provider: A Claude Code session, a Codex thread and a Cursor conversation are not exactly the same unit. Compare a provider against itself over time rather than against another provider.

  • Empty is not zero: OpenAI Platform publishes no session count.

Did this answer your question?