Build an AI Voice File That Actually Sounds Like You
Most marketers approach AI voice training the same way: feed Claude a bunch of your writing and say 'write like me.' Ruben Hassid tried something radically different. Instead of teaching through imitation, he built a voice file through negation — defining not what his voice is, but what it isn't. The result is a single markdown document that makes Claude sound genuinely, uncannily like him. Read the original here.
Start With 47 Minutes of Radical Honesty
Hassid's process begins with what he calls a "voice excavation" — a structured self-interview where you record yourself talking about your work, your opinions, your pet peeves about marketing writing. The key insight: don't try to sound polished. The raw, unfiltered version of how you actually think is exactly what Claude needs.
Build the Anti-Voice First
This is where it gets counterintuitive. Before writing a single word about what your voice sounds like, Hassid documents everything it doesn't sound like. "Never use the word 'leverage.' Never start a paragraph with 'In today's fast-paced world.' Never use more than one exclamation point per piece." These constraints, he argues, do more work than any positive description ever could.
The Markdown File That Changes Everything
The final voice file is a structured markdown document with sections for tone, vocabulary constraints, sentence structure preferences, and what Hassid calls "opinion anchors" — strong takes that color everything you write. Upload it to a Claude Project, and every conversation starts with Claude already knowing how you think.
What makes this approach so effective is its specificity. Instead of vague instructions like "be conversational" or "sound professional," the negation-based voice file gives Claude hundreds of concrete constraints that collectively produce writing that feels genuinely human — because it's built from genuinely human preferences.
“Stop telling AI what you sound like. Start telling it what you'd never say. The constraints do more work than the descriptions ever could.”
— Ruben HassidWhat Ruben Hassid demonstrates here is something every marketing team can learn from. The best AI workflows are not about replacing your voice or your process. They are about finding the right framework that lets AI meet you where you are, and then building on it over time.
Ready to try this kind of thinking in your own workflow? These step-by-step recipes will walk you through related techniques you can start using today.
