A content team is really five jobs wearing a trench coat. Someone finds the ideas. Someone hooks people in the first line. Someone turns a messy outline into a draft. Someone figures out what the audience actually cares about. And someone reads it all back and makes it sound like a human wrote it. Most solo creators try to do all five with willpower and a blinking cursor.
You don't need to hire that team. You need five prompts that each do one of those jobs well. Below are the patterns I actually reach for, with a template you can paste today and a tip to stop the output from reading like a press release written by a robot.
One warning before we start. The reason most AI writing sounds generic is that the prompt was generic. "Write me a blog post about X" gets you sludge. The prompts that work carry your specifics: your examples, your constraints, your weird opinions. Feed the model something real and it hands you something usable.
1. The repurposing prompt
One good piece of content is usually ten pieces you haven't published yet. That webinar, that long email, that Slack rant you wrote to a coworker. This prompt cracks a single asset open and pulls out everything hiding inside.
What it does: takes one long thing and spins it into short-form posts, without flattening your point into mush.
Here's a piece of content I already published: [paste the full text]. Pull out the 5 strongest standalone ideas in it. For each one, write a short LinkedIn post (under 120 words) that makes that single point. Keep my phrasing where it's sharp. Do not add generic advice that wasn't in the original. Do not use hashtags.
Tip to keep it from sounding generic: add "Do not add anything I didn't say." Left alone, the model pads posts with filler wisdom nobody asked for. Boxing it into your actual ideas keeps the voice yours and the claims true.
2. The hook prompt
People decide whether to keep reading in about one second. If your first line is "In today's world, content is more important than ever," you've already lost. This prompt generates a pile of openers so you can pick the one that doesn't make you cringe.
What it does: writes a batch of first lines in different shapes, so you're choosing instead of staring at a blank page.
Topic: [your topic]. Audience: [who they are and what they already believe]. Write 10 opening lines for a post on this. Use a mix of these shapes: a surprising number, a confession, a contrarian take, a specific tiny scene, and a question that stings a little. No line longer than 15 words. No "imagine if" and no "in a world where."
Tip to keep it from sounding generic: name the openers you hate and ban them by exact phrase. Models default to the same three or four hook clichés. Blacklist them and you force it into fresher territory. Then rewrite the winner in your own words anyway.
3. The outline-to-draft prompt
Drafting from scratch is where most people stall. Drafting from a solid outline is fast. So do the thinking first as bullets, then let the model connect them into prose while you stay in charge of the argument.
What it does: turns your skeleton into a real draft without inventing points you never made.
Here's my outline: [paste bullets, one line per point]. Write this as a draft. Follow my order exactly. Do not add new sections or new claims. Keep paragraphs short (2 to 4 sentences). Write like you're explaining it to a smart friend, not presenting to a board. Where I left a bullet thin, flag it with [NEEDS MORE] instead of making something up.
Tip to keep it from sounding generic: that [NEEDS MORE] flag is the whole trick. It stops the model from filling gaps with confident nonsense, and it tells you exactly where you still owe the reader a real thought. Fill those in yourself.
4. The audience research prompt
You can't write for people you don't understand. Before you draft anything, it helps to map what your reader is stuck on, what they've already tried, and what words they use when they complain about it. The model won't replace real conversations, but it's a strong first pass.
What it does: builds a rough profile of your reader's problems and language so your content actually lands where they are.
My audience: [describe them in one or two sentences]. They want to [goal]. List: (1) the 5 biggest frustrations they have on the way to that goal, (2) the things they've probably already tried that didn't work, and (3) the exact phrases they'd type into Google or say to a friend about this. Be specific and concrete. No corporate jargon.
Here's how to use what comes back:
- Steal the frustrations for your subheads. They're the questions readers came to answer.
- Steal the "already tried" list to show you get it before you pitch anything.
- Steal the exact phrases and use them word for word. That's how you sound like an insider instead of a brochure.
Tip to keep it from sounding generic: pressure-test the output against one real customer email or support ticket. If the model's guesses don't match how a live human actually talks, throw them out and paste the real quote instead.
5. The voice-matching editor prompt
This is the one that separates content that sounds like you from content that sounds like everyone. You give the model a sample of your writing and tell it to edit toward that, not toward some polished neutral default.
What it does: rewrites a draft to match your actual voice, using your real writing as the reference.
Here are 3 paragraphs I wrote, in my natural voice: [paste your writing]. Now edit the draft below to match that voice: sentence rhythm, word choices, how formal or casual I am, how I use humor. Keep the meaning intact. Cut anything that sounds like AI or marketing copy. If a sentence sounds nothing like me, rewrite it. Draft: [paste draft].
Tip to keep it from sounding generic: give it a "sounds like me" sample and a "does NOT sound like me" sample. Contrast teaches faster than description. Paste a paragraph you'd never write next to one you would, and the model finally understands the line you're drawing.
Put them in order and they compound
Used one at a time, each prompt saves you twenty minutes. Chained together, they replace a workflow: research the audience, draft from your outline, write the hook, edit to your voice, then repurpose the finished thing into a week of posts. That's the loop a small team runs, and now it runs on your laptop.
If you'd rather not build and tweak these from scratch, a pack like The Grand Prompt Grimoire collects patterns like these so you can grab one and go. Either way, the principle holds: the prompt is only as good as the specifics you put in it.
Keep your ideas, your examples, your voice in every prompt. The AI handles the typing. You handle the thinking. That's the whole trade, and it's a good one.