Describe what you want to write about, who the audience is, and what you want the reader to take away. Grok will propose a structured outline with section headings, key arguments, and a logical flow. You can rearrange, add, or cut sections before any drafting begins.
Example prompt
Write a 2,000-word blog post on why most companies get RAG wrong. Target audience is technical leaders evaluating AI infrastructure. Tone should be direct and opinionated, backed by concrete examples. Cover chunking mistakes, embedding model selection, and evaluation gaps.
Grok writes the full draft based on your approved outline. You can iterate on any section — ask it to sharpen the argument, add a real-world example, cut jargon, make a section more technical, or soften the tone. Each revision is instant and preserves the overall flow.
Example prompt
The chunking section is too abstract. Add a concrete before/after example showing how naive 512-token chunks break a legal contract versus semantic chunking. Also, the intro is too soft — start with the stat about RAG accuracy in production.
Once the content is solid, Grok does a final pass — tightening prose, fixing transitions, checking consistency, and formatting for your publishing platform. It can also generate metadata: title variants, SEO descriptions, social snippets, and pull quotes.
Example prompt
Final pass: tighten every sentence, kill any remaining filler words, and make sure the transitions between sections are clean. Then give me 3 title options, a meta description under 155 characters, and a Twitter thread version of the key points.