How Flyletter is Engineered to Write in Your Voice
AI stops sounding like you after the first paragraph because of structural overload. All Flyletter agents follow a three-layer architecture that fixes it.
You paste your best writing samples into an AI chatbot. You ask it to write in your voice, and the first paragraph sounds good enough.
But then you write a second, and a third.
Before long the rhythm flattens. The word choices become corporate. By the end of the piece the thing sounds like every other draft ever written by AI, and you assume the model just can't write in your voice.
But you're wrong. AI can write in your voice, if you do it right.
What's probably happening is that you're giving AI a little context on who you are and who you write for, then having it read a bunch of old writing samples and execute on a set of writing instructions. Maybe you have a list of AI patterns and phrases you forbid.
No wonder it can't write like you.
You're basically cramming who you are, what you sound like, who you're writing to, and what you're writing into a set of directives that confuses the AI. Compliance collapses under the load, and it breaks most of the prompt engineering best practices I learned building Flyletter's agents.
So the fact that you can't get AI to write like you isn't a model failure. It's a structural failure.
The fix is to engineer a three-tier prompt architecture that sets a quality floor, captures your brand voice, and then instructs the agent on what to do with it.
That's how I built Flyletter, and I'm opening up the kimono to show you exactly how it's constructed so it always sounds like you.
The Three-tier Brand Voice System
Flyletter is built on a three-layer prompt architecture that every agent follows:
- Universal writing standards: All agents start with a set of Universal Writing Standards that serve as the base layer. The Universal Writing Standards is the quality floor that teaches the agents how not to write like AI with banned terms, patterns, and phrases.
- Your brand voice profile: On top of the quality floor sits your brand voice profile that teaches each Flyletter agent about your brand, how you write, and who you're writing for (all built from just three writing samples). Each Flyletter agent gets only what it needs from your brand profile to complete its task.
- The agent instructions: The third layer are the agent instructions that tell each agent how to read your brand profile and how to apply it to their specific task. For example, the draft agent gets instructions on how to turn a detailed outline into a draft newsletter issue, while the social media agents get instructions on how to write banger hooks in your voice.
That's the entire architecture. The rest of this article is me opening up each layer to show you how it's built. Feel free to try and do this yourself, and I encourage you to try Flyletter and compare the output against your own.
1. Universal Writing Standards
The quality floor is a set of Universal Writing Standards that all Flyletter agents follow so they don't write like AI. This layer doesn't need to know anything about your brand yet. It just needs to know what good writing looks like in general, as well as what generic AI sounds like so it can strip it out.
These Universal Writing Standards are a list of instructions split into two broad halves. The first part shows what good writing looks like. The second part calls out the AI tells and forbids them.
What Good Writing Looks Like
The first half sets the minimum standard every output should clear before it tries to assume your voice, which includes but is not limited to:
- Sentence craft: Vary length on purpose. Short punches, medium carries, and long builds.
- Concision: Every word earns its place. One strong word beats a weak word plus a modifier.
- Paragraph discipline and rhythm: One idea per paragraph. A standalone one-sentence paragraph is powerful when a thought earns its own space.
- Formatting discipline: Every bold, header, and structural element serves the content.
- The anti-fabrication standard: A fabricated story destroys your credibility the instant a reader recognizes it.
- The Human Test: Before anything ships, ask, "would a busy, smart human write this way?"
That's the ground truth. The standard that all Flyletter output must adhere to.
Top AI Tells and Giveaways
Even though the AI knows the mechanics of good writing, it still needs to know what bad writing sounds like and how to avoid the dead AI giveaways:
- The three top giveaways: Stage direction ("here's the thing"), em dashes (sad, because I love em dashes), and the colon for dramatic pause.
- The pattern families underneath: Contrast reveals ("This isn't X. It's Y."), sweeping scene-setters ("In today's landscape"), dramatic imperatives ("Picture this:"), and more.
- The forbidden terms and phrases: The named words that flag AI on sight. Things like "unlock" and "unleash."
That's the floor. It ensures at minimum that what the AI writes won't sound like it was written like AI. It still won't sound like you, though, which brings us to the next layer.
2. Your Brand Voice Profile
Most AI tools that say they write in your voice give you limited options, like "friendly but professional," and call the job done. Not Flyletter.
Flyletter's brand voice profile is built from just three writing samples and teaches Flyletter's agents how to write like you across six key categories:
- Audience focus: your target audience and unique differentiator.
- Brand positioning: who you are, who you write for, and what you stand for.
- Writing style: the overall sound and structure of your writing.
- Signature elements: elements that make your writing uniquely yours.
- Forbidden elements: words and patterns you never use.
- Visual style: how your images and visuals should look and feel.
Each Flyletter agent gets the exact pieces of the brand profile it needs to write like you. For example, the draft agent may get the full voice kit, while the topic suggestion agent may only get your audience and brand positioning.

But a robust brand profile alone doesn't result in good writing output. Something has to understand and apply it.
3. The Agent Instructions
So you have the quality floor in a set of Universal Writing Standards, a brand voice profile the AI can read from, and now you need the actual agent instructions.
Every Flyletter agent has instructions on exactly how to understand your brand profile and apply it to its specific task. While the agent instructions are all role-specific, Flyletter agents follow a few core principles:
The Character Principle
The information in your brand profile is evidence of how you sound, not a checklist to deploy line by line.
If the agent treats every field in the profile as a box to check in every sentence or paragraph, you get a draft that over-signals every trait at once and reads like an impression of you.
A checklist produces a caricature, but reading it as evidence to apply when appropriate is what gets AI to sound more naturally like you.
Reading fields as evidence means the engine decides which patterns matter for the specific piece and leaves the rest quiet. Your fragment habit shows up where a fragment lands, not on a set schedule.
Per-Agent Slicing
The same brand profile gets sliced and delivered to each agent differently depending on the job so they don't get overwhelmed with information.
Long-form writing agents get the full voice profile, where paragraph rhythm and argument architecture actually matter. But agents like the subject line agent get different slices of the persona.
For the subject line agent, paragraph rhythm doesn't matter, but context on your brand's unique differentiator may help it write more engaging subject lines. Force the full-voice instruction set onto an agent trying to write six words and you confuse the model and degrade its output.
Voice-neutral Instruction Language
Flyletter's agents have to write naturally across different use cases and archetypes. The instructions themselves can't therefore secretly bias the output toward one archetype.
Write your directions in punchy, clipped, imperative language and you quietly push every voice toward punchy, even the writers who ramble beautifully and should. For that reason, Flyletter uses voice-neutral language for its instructions so it doesn't distract the model.
When the Three Tiers Disagree
Sometimes the three tiers will disagree. For example, the Universal Writing Standards strips an AI tell that the brand profile insists is a signature element of yours. When that happens, something has to win, and the layers outrank each other in a reverse order:
- Safety: Non-negotiable, always first.
- The writer's voice: Your patterns outrank the machine's defaults.
- The agent's task rules: The job-specific instructions for this task.
- The universal floor: The generic-tell cleanup.
Notice voice sits above both task rules and the floor. That's on purpose. If the floor outranked your voice, it would flatten the exact quirks that make you sound like you.
Bottom Line
Flyletter writes in your voice thanks to a three-tier prompt architecture. First, it follows a set of Universal Writing Standards. Then, it learns your brand voice, and third, it applies its agent instructions.
It's the best way I've found to architect agents to write in your voice. Otherwise, there's too much instruction and cognitive load for the AI to follow.
