how to maintain brand voice (and stop sounding like a committee)
your substack sounds like one person. your linkedin, another. your blog, a third. it's not a people problem. it's a system problem. one that has a fix.
you hire a writer. the first draft lands. it's… fine. grammarly fixed the commas. the points are all there. but the voice is gone. it sounds like something a committee would write. not you.
so you spend an hour rewriting it. you do it again for the next blog post. and again for the newsletter. soon enough, you're the company's final, exhausted voice filter. that doesn't scale.
the anatomy of voice drift
voice drift isn't a writer problem. it's a systems problem. it happens when you have multiple writers, editors, and ai tools all working from a static style guide. there's no single, measurable source of truth. you don't have bad writers. you have a great system for producing average content.
why static guides don't work
your brand voice guide is probably a notion doc full of adjectives. 'bold.' 'witty.' 'insightful.' that’s a wish list, not a tool. it can’t tell a new writer that your 'boldness' is a specific ratio of declarative to interrogative sentences. or that your 'wit' is a pattern of subverting expectations with a short fragment. it describes the effect, not the cause.
the more adjectives in a brand voice guide, the less consistent the writing. we see it all the time. writers get paralyzed trying to perform 'witty' instead of matching the structural patterns that produce it.
this is the failure mode. a freelancer reads your 20-page guide, tries to be 'bold,' and just ends up overusing imperative verbs. you get a caricature of your voice, not the real thing. the guide described the city but handed them a broken compass.
the first signals of drift
the first signal is a collapse in structure. our analysis of 500+ voice profiles shows that writers using a static pdf guide score 55% lower on voice in their first month than those using a dynamic profile. their sentence-length variation flattens by nearly 70%. their use of your brand's key transition phrases disappears by the third post.
think about the rhythm in a justin welsh post. short, staccato lines. then a longer, multi-clause explanation. a new writer or an ai draft smooths this out. you get a flat line of 15-word sentences. the meaning is there. the voice is gone.
how to maintain voice without going insane
you replace the static guide with a dynamic voice profile. one that can score new content instantly. this becomes the single source of truth. for your head of content, your new marketing hire, your freelancer. everyone gets the same objective feedback. the process stops being subjective and becomes measurable.
npm i -g @holdyourvoice/hyva practical workflow that actually works
the workflow is simple. no new meetings required. it has three parts:
- build your voice profile from your best work. not adjectives. find your 5-10 most 'on-brand' pieces. use that corpus to build a measurable profile.
- score every single draft. before a human ever sees it, run the content through the voice profile. it doesn't matter if it's from a junior writer or jasper. you get an instant, objective score from 0 to 100.
- edit for patterns, not just words. the score gives you specific feedback. not 'make this sound more like us.' more like, 'increase sentence length variation by 20% and add back one of our core transition phrases.' that's feedback a writer can use.
this workflow turns voice from a mystical art into a repeatable craft. you're no longer the bottleneck. your team can finally sound like the best version of your brand, without you having to rewrite everything.
get started for free — install hyv, paste the command in your terminal, and run onboarding in seconds.
npm i -g @holdyourvoice/hyvget started for free — install hyv, paste the command in your terminal, and run onboarding in seconds.
npm i -g @holdyourvoice/hyv
writes about brand voice, ai writing patterns, and the craft of sounding like yourself. built hold your voice after watching his own voice flatten across six months of heavy ai drafts.
co-written with ai as sidekick. shashank drafted the observations; the ai pressure-tested the structural claims. if something reads too smooth, that's the ai's fault.