# How is Claira different from TAR?

> TAR learns from reviewer decisions to rank documents. Claira reads documents against your instructions and explains its calls, from the first document.

Technology-assisted review (TAR) trains a model on documents your reviewers code, then ranks the rest by how likely they are to be relevant. It needs a seed set, and its output is a score rather than a reason.

Claira reads documents against instructions you write in plain words and can give a call and its reasoning from the first document, with no training round. You can change the instructions and rerun them at any time, and calls can be checked one by one.

The two can work together: many teams use Claira for the first pass and keep their validation methods for the final set.