The tool has two modes. Light is a single fast pass: the points go in as one request and come back as connected prose, no sections, no highlighting — the same way the site's other text tools work. Best builds a plan and breaks the result into sections: each one can be edited on its own, and material the model added that wasn't in the original points gets highlighted, so it's clear what's worth double-checking first. Where the model tends to go wrong when expanding, and how to check the result, is covered in how an AI text expander works.
Settings in both modes: an optional goal for the text; length as a multiple of the input (×2/×3/×5) or an exact word count; format (continuous prose, an article with subheadings, a social post, a letter, a speech, a product description); tone from six options and audience from four; what to add — examples, arguments, transitions, an intro and conclusion, questions to the reader, a short takeaways list (several can be picked at once); toggles for "stick to the points only" (bars the model from inventing facts or figures that weren't in the source) and "keep the original order"; point of view, output structure, keywords and a style sample for reference; and a control for how much freedom the model takes.
In best mode, the result can be refined without regenerating the whole text: expand, shorten, rewrite, or change the tone of a single section, plus footer buttons for "longer," "shorter," "add an intro," "add a conclusion," and "continue" for the whole piece. Expand your points into text — three free runs a day are available without signing in.