The detail
Blog content is a standard unit of AI writing work. Generating a 1,200-word post means paying for ~1,600 output tokens — and output rates are typically 3-8x input rates, so generation costs meaningfully more than reading the same text.
Editing workflows flip the ratio: sending a full draft for revision costs ~1,600 input tokens plus the revised output. A full rewrite therefore bills roughly double the post length, while targeted section edits cost a fraction of that.
Markdown structure (headings, links, lists, code snippets) tokenizes heavier than plain prose. A technical post with code examples can run 1.5-2x the token count its word count suggests, so budget accordingly for developer-focused content.