Skip to main content
Table of Contents

Improving AI Attribution Accuracy and Coverage

Best practices for improving AI attribution accuracy and coverage in LinearB using commit metadata, branch signals, contributor mapping, and frequent commits.

Steven Silverstone
Updated by Steven Silverstone

Overview

AI attribution accuracy in LinearB depends on the quality of the signals available from Git activity and connected AI tools. Use this article to improve both direct attribution and heuristic coverage in AI Analytics and related reporting.

For background on attribution logic, see How LinearB Calculates AI Attribution.

You do not need every signal for attribution to work. Direct Git metadata provides the highest confidence, while tool usage and timing signals help improve coverage when direct metadata is missing.

1. Use Co-Author Commit Metadata Where Available

When an AI tool adds a Co-authored-by trailer to a commit message, LinearB can identify the commit as AI-assisted directly from Git. This is the strongest attribution signal because it is attached to the commit itself.

For other AI coding tools, check the vendor documentation for commit attribution or co-author settings. If a tool does not support this natively, use the same Git hook or commit helper pattern.

Examples: Claude Code includes a co-author byline by default. GitHub Copilot in VS Code can add one when git.addAICoAuthor is configured as chatAndAgent or all. Codex can add a co-author trailer through commit attribution configuration. For Antigravity workflows, use a team-owned Git hook or commit helper when the tool does not add a specific co-author trailer.
Co-authored-by: AI Tool Name <ai-tool@example.com>
Use an email address your organization recognizes. Avoid placeholder addresses in production workflows.

2. Keep AI Branch Signals Distinct

Some AI coding tools create branches with recognizable prefixes. When those prefixes are specific to AI-assisted work, LinearB can use them as an additional signal for attribution.

To preserve branch-based coverage:

  • Keep AI-specific branch prefixes enabled when your tool supports them.
  • Avoid changing AI branch prefixes to generic values such as feature/, bugfix/, or dev/.
  • Use branch naming consistently across teams and repositories.
Examples: For OpenAI Codex, use Git → Branch prefix in Codex settings to keep a distinct prefix such as codex/. GitHub Copilot coding agent branches begin with copilot/.

Branch prefixes are a supporting signal. Co-author commit metadata is still more reliable when both are available.

3. Map AI Tool Users to Contributors

AI tool integrations can create bot or service account users that represent AI activity. To include that activity in team-level and user-level reporting, make sure those users are mapped or merged with the correct contributors in LinearB.

To review unmapped AI tool users:

  1. Go to Settings → Company Settings → Users & Teams → Users.
  2. Click Add Filter.
  3. Select AI Tool Connection.
  4. Select the relevant AI tool, such as Claude Code, Cursor, GitHub Copilot, or Amazon Q.
  5. Select Doesn’t have account.
  6. Review the filtered users and merge or map them to the correct LinearB contributors where needed.

For step-by-step instructions, see Merge Contributor Accounts.

Review contributor mapping when you add a new AI integration. Unmapped AI tool users can cause activity to be missing from team-scoped or user-scoped reports.

4. Commit Frequently When Using AI Tools

When direct Git metadata is not available, LinearB can use heuristic attribution by correlating AI tool usage with Git activity. This works best when commits happen close to the time the AI tool was used.

Encourage developers to commit at least once per day when actively using AI tools. Smaller, more frequent commis improve timing correlation and reduce the chance that AI-assisted work is missed.

This is especially important for tools or workflows that do not add co-author commit metadata.

Summary

Action

Signal Type

Impact

Use co-author commit metadata

Git-based direct signal

Highest-confidence commit attribution

Keep AI branch signals distinct

Git-based supporting signal

Improves coverage for agentic workflows

Map AI tool users to contributors

User mapping

Required for accurate team and user reporting

Commit frequently

Heuristic timing signal

Improves coverage when direct signals are absent

How did we do?

How LinearB Calculates AI Attribution

Contact