Brand Voice Drift
The measurable divergence between a brand's stated voice and its produced content over time, especially across channels.
What drift looks like
Brand voice drift happens slowly. The founder writes the homepage. A contractor writes the about page. An agency posts on LinkedIn. The intern handles Instagram. Three months later, your brand sounds like four different brands.
Drift is measurable. The whystrohm-voice-scorer skill computes a drift score between any two content sources by comparing vocabulary, cadence, structure, and tone. Below 70 means real divergence. Below 50 means the channel reads as a different brand.
Why drift is worse with AI
AI tools amplify drift. Every team member uses a different AI tool with a different prompt. The same brand gets six different voice profiles depending on who's generating. Without a shared, enforceable voice config, AI doesn't reduce drift — it accelerates it.
How to prevent drift
Three controls actually work:
- Extract the voice once. A canonical brand-config.json that every AI tool reads from.
- Enforce at generation. The voice profile becomes a constraint, not a suggestion.
- Score after generation. Every output gets scored against the profile before publishing.
This is the content firewall pattern.
Related Terms
- AI SlopGeneric, voiceless content produced by AI systems without brand governance. The opposite of deterministic content infrastructure.
- Stylometric FingerprintA quantitative profile of a writer's or brand's voice across cadence, vocabulary, structure, and rhetorical patterns. Reproducible and enforceable in code.
- Deterministic ContentContent produced by systems that yield the same brand-aligned output for the same input, every time. Opposite of probabilistic AI generation.
Skills That Address This