Brand voice — what it is, and why it's not "tone"
A brand voice is a set of choices about *how* you sound that compound across every touchpoint. Done well, voice creates trust before any data does. Done poorly (or inconsistently), it actively destroys trust.
The plain-English version
If you read three random ads from the same brand and they sound like they were written by three different people, the brand doesn't have a voice yet — it has a vocabulary.
A brand voice is the answer to: "if our brand were a person, how would they speak?" Sage Vitality (Eudaven's brand voice) is considered, unhurried, clinically warm, and respectful. Every word choice across every ad has to pass through that filter.
This is not the same as tone. Tone changes with context (you can be playful in a welcome email and serious in a recall notice). Voice is constant — the underlying personality. Tone is the inflection; voice is the speaker.
Why voice matters for GLP-1 specifically
The GLP-1 vertical is swimming in voice violations. Compounded competitors lean on hype ("transform your body in 30 days!"), urgency ("limited time!"), and clinical-sounding-but-empty language ("doctor-approved!"). All of those read as red flags to sophisticated patients (Rachel, Jess), and the FDA + platforms have made them increasingly risky.
Eudaven's voice was designed in part to be the opposite of the generic compounded competitor. When a Rachel-type patient reads a Sage Vitality ad after reading three hype-driven competitor ads, the calmness itself is a differentiator.
The Sage Vitality voice attributes
The voice is captured in data/kb/brand/sage-vitality.md. The
operational essentials:
- Considered, not breathless. No exclamation points. No superlatives. Phrases like "transformative," "incredible," "amazing" are banned.
- Unhurried, not urgent. No "limited time," "act now," "while supplies last." Urgency is treated as dignity violation under GLP1-016 because the urgency is manufactured rather than real.
- Clinically warm. Speak like a doctor who actually listens. "Your clinician will determine the right starting dose" — not "we'll get you a prescription fast."
- Respectful of the reader's body. No "fix your body," "transform your shape," "before you're stuck this way." The reader's body isn't a problem to be solved.
- Specific over generic. Name the medication, the dosing approach, the pharmacy partner. Generic claims read as evasive.
Voice rules that have specific compliance overlap
Several voice rules double as compliance rules. Understanding both sides clarifies why the voice rule exists:
- No hype ↔ FTC substantiation. Hype words ("transformative," "guaranteed") often imply efficacy that can't be substantiated.
- No urgency ↔ GLP1-016 dark-pattern rule. Manufactured urgency is a regulatory issue, not just a voice issue.
- No "fix your body" ↔ GLP1-011 dignity violation. The voice rule and the compliance rule are the same rule expressed at different levels.
When you write copy that feels right for voice but trips compliance, the fix is often a voice fix that incidentally resolves the compliance issue. They're aligned by design.
How voice compounds
Voice's effect compounds in a way single-ad metrics can't measure:
- One ad that sounds calm and credible doesn't move the needle.
- Ten ads over six months that consistently sound calm and credible build a pattern in the reader's mind: "this brand is different."
- A landing page, an email sequence, a clinician's intake message, an SMS reminder, and a customer-service reply that all sound like the same person built trust the competition can't fake.
This is why voice consistency matters more than any individual ad's performance. A great ad in the wrong voice destroys six months of trust-building.
How the pipeline enforces voice
The brand scorer node (app/pipelines/nodes/brand_scorer.py)
runs after copy_polish and scores variants against the voice
attributes from data/kb/brand/sage-vitality.md. The score is
returned as one of the variant version fields; below a threshold,
the variant is flagged for human review.
The compliance node and the brand scorer are separate by design. Compliance asks "is this legal/safe?" Brand scoring asks "is this Eudaven?" A variant can pass compliance and fail brand voice — and we treat that as a real failure, not a nice-to-have.
What to remember
- Voice is constant, tone varies.
- Voice violations are usually compliance violations too.
- Calm + specific beats hype + generic, especially with our personas.
- One off-brand ad costs six months of trust-building.
What to do next
- Read Personas if you haven't — voice attributes resonate differently with each persona.
- Look at three real Sage Vitality variants and a competitor's ads side by side. The voice difference is the point.
- Skim
data/kb/brand/sage-vitality.mdfor the full voice definition.