Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →To track AI-generated code across repositories, combine four separate records: coding-tool usage telemetry, changes the tool attributes to AI, pull-request activity, and provenance that links a commit to an agent session. They answer different questions. None, by itself, reveals every AI-assisted edit or proves that AI caused a productivity gain.
What are you trying to measure?
Start with the question, because “AI contribution” can mean several things. Adoption is about who uses an assistant and how often. Contribution is about changes the product identifies as user-initiated or agent-initiated. Repository throughput concerns pull requests and their progress. Provenance connects a change to the session that produced it.
| Signal | What it can tell you | What it cannot establish on its own |
|---|---|---|
| Tool usage telemetry | Who used a product and reported activity such as suggestions or accepted suggestions, depending on the available metric. | Which resulting code reached a repository or whether it improved delivery. |
| Tool-attributed code changes | Changes the coding product classifies as user-initiated or agent-initiated, including line counts where supported. | All AI assistance across tools, or the value, quality, or authorship of every line. |
| Pull-request activity | Repository-level PR events, such as PRs created by an agent or reviewed by an AI tool, where reported. | A complete count of AI-generated code or the full history of assistant use. |
| Session provenance | A trace from a commit or change to an identifiable agent session, when the product provides one. | A universal cross-vendor record for every assistant and every AI-assisted edit. |
What GitHub Copilot reporting can show
For a GitHub Copilot estate, GitHub documents usage metrics through dashboards, APIs, and NDJSON exports, with enterprise-, organization-, repository-, and user-level reporting. Available report shapes vary by scope and purpose; check the relevant metric definition before combining records. See GitHub Copilot usage metrics and the data available in Copilot usage metrics.
Usage telemetry
Usage metrics help answer adoption questions, but they are not a ledger of repository changes. GitHub says most metrics depend on client-side IDE telemetry. Settings and supported IDE or plugin versions affect coverage, and richer telemetry may be unavailable for some measures. Treat missing data as unknown rather than as proof that no assistant use occurred. GitHub’s usage-metrics documentation describes these dependencies.
Recommended Free Tools
#1 Best Overall
Code-generation attribution
GitHub’s code-generation reporting distinguishes user-initiated from agent-initiated changes and can report lines added or deleted. GitHub characterizes these lines-of-code measures as directional: they describe Copilot output across supported completions, chat, and agent features, not a universal measure of AI authorship or engineering value. Coverage depends on supported IDE and plugin versions. Consult GitHub’s Lines of Code metrics documentation for definitions and limitations.
Do not equate a suggested line with a line that was ultimately committed, or a line count with quality or productivity. Preserve the metric’s exact definition—such as suggestions, accepted suggestions, added lines, or deleted lines—when storing or displaying it.
Pull-request activity
Repository-level PR reporting describes daily repository activity. It can include PRs created by Copilot cloud agent or reviewed by Copilot code review, but it is not a universal count of AI-generated lines. A repository with no activity for the requested day is omitted from that report, so absence from the output is not the same as an explicit zero. The report details are documented in Data available in Copilot usage metrics.
Agent-session provenance
For Copilot cloud agent, GitHub documents a more direct provenance trail: Copilot is the commit author, the person who started the task is listed as co-author, and the commit message links to the session logs. This makes it possible to inspect that agent workflow; it is not a general mechanism for every coding assistant. See Managing agent sessions.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Rank #3
Build a repository-wide reporting process
- Choose the question and metric. Decide whether you need adoption, attributed code changes, PR flow, or session provenance. Do not label one category as another.
- Set a consistent reporting scope. For Copilot, select the enterprise, organization, repository, or user-level data that matches the question. For a portfolio view, export the available records or use the API, then join them to a stable repository inventory. Preserve the original scope and reporting date for each record.
- Record definitions alongside values. Store the provider and product surface, reporting window, repository identifier, user or agent attribution, metric type, and known telemetry gaps. A value for accepted suggestions is not comparable to a value for added lines or PRs.
- Keep attribution rules consistent. Organization and enterprise totals can differ because of deduplication and attribution timing. Do not add or compare totals from different scopes as if they were equivalent; use the rules documented for the specific report in GitHub’s metrics data reference.
- Retain explicit provenance where available. Keep agent identity and session links with the relevant commit or PR metadata so reviewers can inspect the trace. When a tool does not supply such evidence, record attribution as unknown or tool-reported rather than inferring it from code style.
- Pair activity with outcomes when evaluating impact. Compare AI activity with trusted team measures such as review and merge flow. A relationship between adoption and PR output, including the cohort reporting described in GitHub’s metrics overview, is not by itself proof that adoption caused a change in productivity.
Can you tell whether code was written by AI?
Only sometimes, and with different levels of confidence. A product’s explicit attribution or an agent-session link is evidence about that product’s workflow. Usage totals are broader but do not identify particular committed lines. A classifier that looks for behavioral fingerprints is a different kind of evidence and should not be treated as audit-quality provenance.
A 2026 study by Taher A. Ghaleb, “Fingerprinting AI Coding Agents on GitHub,” analyzed 33,580 pull requests from five agents and reported a 97.2% F1-score for identifying agents in that dataset. That is a study result on the paper’s analyzed PRs, not a guarantee of accuracy for another codebase, an individual change, or every AI assistant. See the paper.
The available documentation establishes detailed Copilot reporting and cloud-agent provenance, but not a shared attribution format across all coding assistants and repository platforms. Do not claim that every AI-assisted edit can be detected after the fact.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What to compare when choosing a tracking approach
- Product coverage: Which assistants, agent modes, and product surfaces are included?
- Scope and granularity: Can you report by organization, repository, user, agent, or session?
- What is counted: Suggestions, accepted suggestions, lines, commits, PRs, or outcomes are distinct measures.
- Coverage limits: What client telemetry, IDE versions, or plugins are required, and how are missing records represented?
- Portfolio reporting: Are exports or APIs available, and can records be joined reliably to your repository inventory?
- Auditability: Does the system preserve explicit agent identity and a session trail that reviewers can inspect?
- Retention and definitions: How long are records available, and are metric definitions and reporting windows clear enough to compare over time?
These checks help distinguish a useful operational dashboard from evidence of authorship. A dashboard can describe reported activity; a traceable session can document a particular agent workflow. Neither should be presented as more comprehensive than its coverage allows.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
