Why your AI writing sounds generic (and the 3-layer fix)

Hey,

Most people quit on AI writing for the same reason: it sounds generic. Bland. Like a press release nobody asked for. So they decide “AI can’t write” and go back to doing it all themselves.

Here’s the thing, it’s almost never the model. The same ChatGPT that gave you that lifeless paragraph can write something genuinely good. The difference is what you put in front of it. Today I’ll give you the exact structure I use, the “context sandwich,” so your AI stops sounding like AI.

IN THIS EMAIL

» Why “generic” happens (it’s a context problem) » The context sandwich, the 3 layers that fix it » A fill-in-the-blank template to steal

🤖 Why everything comes out bland

When you type “write a post about morning routines,” the model has nothing to work with except the average of everything ever written about morning routines. So it gives you exactly that, the average. Generic in, generic out.

A good writer doesn’t just know the task. They know who they’re talking to, what’s actually true from experience, and the voice it should be in. You have all of that in your head. The reason your output is bland is that you never handed any of it over. You gave the model a task and expected it to read your mind.

The fix isn’t a magic phrase. It’s giving the model the same context a real writer would need before they typed a word.

🥪 The context sandwich

Think of every strong prompt as three layers.

Top layer, the frame. Who is the model being, who is it for, and what’s the goal? “You’re a copywriter who’s written for creators with small, loyal audiences. You’re writing for first-time founders who are skeptical of hype. The goal is to make them feel understood, then give them one thing to try.” This one move alone kills 50% of the blandness.

Middle layer, your raw material. This is the part everyone skips, and it’s the most important. Paste your actual opinion, a real example, the messy bullet points, a thing a client said, the way you’d explain it out loud. Even better, paste two or three lines of your own writing and say “match this voice.” The model can’t sound like you if it’s never seen you. Give it something real to chew on.

Bottom layer, the rules. Format and constraints. Length, structure, what to avoid. “Short paragraphs. No buzzwords like ‘leverage’ or ‘unlock.’ No emoji. End with one question.” Constraints don’t limit the model, they focus it.

Frame on top, your raw material in the middle, rules on the bottom. That’s the sandwich. Miss the middle layer and you’ve just got two pieces of bread, which is exactly what generic AI writing is.

🔒 Steal this template

Keep this somewhere and fill in the blanks before you write anything:

ROLE: You’re a [type of writer] who writes for [audience]. The goal of this piece is to [goal]. VOICE: Match the tone of these lines of mine: [paste 2 to 3 sentences you’ve written]. RAW MATERIAL: Here’s what I actually think and the example I want to use: [dump your real take]. RULES: [length], short paragraphs, no buzzwords, no fluff, [anything to avoid]. TASK: Now write [the thing].

The first time you use this it’ll feel like more work. It is, for about a week. Then it becomes automatic, and you’ll never go back to one-line prompts, because the output is on a completely different level.

Try it today: take the last thing AI wrote for you that felt flat, rebuild the prompt as a sandwich, and run it again. The same model, a totally different result.

• László

P.S. When my team builds content systems for clients, this is step one, we capture the brand’s real voice and turn it into prompts so the output sounds like them at scale, not like a robot. If you want that for your business, that’s what we do at Growth Lab Studios. Reach out here.

P.P.S. Want the back catalog? Every past issue lives in the newsletter archive.