Why AI Content Sounds Like AI
You can spot AI-written content from a mile away. It's technically correct but somehow hollow. Professional but personality-free. It reads like it was written by someone trying very hard to sound like everyone.
That's because it was.
Generic AI models are trained to produce "good" writing—which means writing that appeals to the average person. The result is aggressively average: competent, safe, forgettable.
"Excited to share..." "I'm thrilled to announce..." "Here's what I learned..." The AI defaults to these patterns because they're statistically common in its training data. They're not wrong. They're just not you.
What a Persona Actually Does
A persona is a set of instructions that shapes how the AI writes. Instead of generating generic professional content, the AI generates content that follows your specific guidelines.
Think of it like briefing a ghostwriter:
Without a persona: "Write a LinkedIn post about remote work."
With a persona: "Write a LinkedIn post about remote work. Be direct and slightly contrarian. Draw from my experience running distributed teams for 5 years. Avoid corporate buzzwords. Use short paragraphs. End with a genuine question, not engagement bait."
Same topic. Completely different output.
Anatomy of a Persona
Postiller personas capture several dimensions of your voice:
Communication Style
How do you structure your thoughts? Options include:
- Direct — State conclusions first, then support them
- Conversational — Write like you're talking to a colleague
- Analytical — Lead with data and evidence
- Storytelling — Open with anecdotes, build to insights
Most people are combinations. "Direct but conversational" or "Analytical with occasional stories."
Tone
What's the emotional register of your writing?
- Professional but warm
- Casual and approachable
- Thoughtful and measured
- Energetic and enthusiastic
- Dry and understated
The AI uses these descriptors to calibrate word choice, sentence structure, and overall feel.
Expertise Areas
What topics do you write about with authority?
- Product strategy
- Engineering leadership
- Early-stage startups
- Marketing analytics
- Whatever your actual domain is
When you generate content touching these areas, the AI writes with more confidence. It assumes you know what you're talking about—because you do.
Phrases You Love
Language patterns that feel natural to you:
- "The thing is..."
- "Here's the uncomfortable truth:"
- "I've seen this pattern before"
- Starting sentences with "And" or "But"
These get woven into generated content, adding your verbal fingerprints.
Phrases to Avoid
Language that makes you cringe:
- "Excited to announce"
- "Thought leader"
- "Synergy" / "leverage" / "circle back"
- "Just wanted to share"
- Excessive hashtags
The AI actively avoids these patterns, preventing the generic AI sound.
Creating Your Persona
The setup process asks questions designed to surface your authentic voice:
How do you typically open posts? Review your past content. Do you start with questions? Bold statements? Stories? Data? Your opening pattern is a strong signal of your style.
What makes you different? Not "what makes you better"—what makes you distinctive? Unusual background? Contrarian views? Specific experience? This shapes how the AI positions your perspective.
Who are you writing for? Your imagined audience affects tone. Writing for senior executives sounds different than writing for early-career peers, even on the same topic.
What do you want people to feel? Informed? Inspired? Challenged? Entertained? The desired emotional response guides the AI's approach.
Multiple Personas for Different Contexts
You're not the same writer everywhere. How you write on LinkedIn differs from Twitter differs from a company blog. Postiller supports multiple personas so you can match voice to context.
Example persona set:
LinkedIn Professional
- Direct, thoughtful, expertise-forward
- Longer posts, complete arguments
- Industry-specific terminology OK
- Avoid: casual language, thread format
Twitter Casual
- Punchy, conversational, slightly irreverent
- Short sentences, strong hooks
- Hot takes welcome
- Avoid: corporate speak, excessive qualification
Company Blog
- We voice, not I voice
- Educational, helpful, authoritative
- Clear structure with headers
- Avoid: personal opinions, controversial takes
When you generate content, select which persona to use. The same source material produces different output for different contexts.
How Personas Affect Generation
The persona isn't just a suggestion—it's woven into the AI's instructions at a system level.
Here's a simplified version of what the AI receives:
SYSTEM PROMPT:
You are writing as [User Name], who has the following voice:
Style: Direct and conversational. States opinions clearly.
Tone: Thoughtful but accessible. Avoids academic language.
Expertise: Product strategy, B2B SaaS, startup scaling.
Patterns to use: Short paragraphs. Occasional "Here's the thing:" openings.
Patterns to avoid: "Excited to share", buzzwords, engagement bait questions.
Generate content that sounds authentically like this person, not like generic AI output.
[USER'S RELEVANT CONTEXT AND NOTES]
Task: Write a LinkedIn post about [topic].
The AI treats these instructions as constraints. It's not just trying to write well—it's trying to write well within your defined parameters.
The Difference Personas Make
Without persona:
I'm excited to share some thoughts on remote work! After years of experience, I've learned that communication is key. Here are my top 3 tips:
- Over-communicate with your team
- Set clear expectations
- Build trust through transparency
What are your best remote work tips? #RemoteWork #Leadership #Future
With a direct, slightly contrarian persona:
Hot take: Most remote work advice is wrong.
"Over-communicate!" they say. But constant Slack messages and status updates create more noise, not more clarity.
The teams I've seen thrive remotely do the opposite. They communicate less frequently but more intentionally. Weekly async updates instead of daily standups. Documentation instead of meetings. Trust instead of surveillance.
The goal isn't maximum communication. It's minimum viable coordination.
Same general topic. Completely different voice, structure, and point of view.
Personas Don't Replace Judgment
Important caveat: personas guide the AI, they don't control it perfectly.
You should still:
Review everything. The AI might occasionally slip into patterns your persona forbids. Quick edit fixes it.
Fact-check claims. If the AI attributes a statistic or experience to you, make sure it's accurate.
Add your actual insight. Generated content is a starting point. Your unique additions make it genuinely valuable.
Evolve your persona. Your voice changes over time. Update your persona when it stops feeling right.
Personas make AI output dramatically better. They don't make human review optional.
Getting Started
If you're not sure how to define your persona, try this:
-
Gather examples. Find 5-10 posts you've written that you're proud of. What patterns do you notice?
-
Ask someone. How would a colleague describe your communication style? Often others see our patterns more clearly than we do.
-
Note your irritations. What writing patterns annoy you? Your "avoid" list often defines your voice more than your "use" list.
-
Start simple. You don't need to fill every field. A basic persona with tone and a few avoid phrases beats an empty one.
-
Iterate. Generate content, see what feels off, refine the persona. It takes a few cycles to dial in.
Your persona gets better as you use it. The AI's output teaches you what your voice actually is.
Why Voice Matters
Content is abundant. Anyone can generate infinite posts about leadership, productivity, marketing, whatever.
What's scarce is perspective. A distinctive voice. A point of view that could only come from a specific person with specific experiences.
Personas help you scale what makes you valuable without losing what makes you distinctive. The AI handles the mechanics of writing. Your voice and your knowledge provide the substance.
That's the combination that produces content worth reading.