Editorial policy.

We choose which repos to test, and the verdict follows what happened in the test.

Stars and popularity are context. They do not move a verdict.

A sponsor can buy a clearly labeled placement, or pay for the work of an evaluation. A sponsor cannot buy a verdict, a ranking, or a change to the evidence.

Paying for an evaluation does not guarantee coverage or a good result. A sponsored repo gets the same test, and its failures are reported the same way.

We disclose conflicts of interest, affiliate relationships, credits a project gave us, and help from maintainers.

Kahtaf Alam, GitTested's editor, approves every review before it is published. He checks that the review names the tested commit, that its claims have evidence behind them, and that failures and limits are included.

Reviews are published under the GitTested name. The editor is accountable for what they say; About says who he is.

AI agents run the tests and draft the review. The editor decides what gets published.

• Testing. An AI coding agent runs the tool on a disposable machine and records the commands and outputs. The editor vets the test design first.
• Writing. An AI agent drafts the review from that record and from research into the project.
• Checking. Separate AI sessions that did not write the draft check its claims against the saved evidence. Their findings go back to the writer.
• Cover art. Cover illustrations are made with an image model from a written concept, and the editor approves them. Screenshots inside a review come from the run.
• Publishing. The editor reads the result and makes the call.