hold your voice for brand voice review
a grammar-clean draft can still sound like it was assembled by three people and a chatbot.
what is hold your voice, and what does it check?
hold your voice checks whether a draft still matches its intended voice, with attention to tone, vocabulary, structure, point of view, and ai-like writing patterns. it gives editors a place to inspect the lines that feel imported from a different company before those lines survive three approval rounds.
the usual failure is quiet. a landing page begins in a direct, practical voice, then halfway through becomes formal because a product marketer added feature copy from notion. a linkedin post starts with a founder’s useful observation, then falls into “in today’s fast-paced environment.” a support article explains a setting plainly until the final section starts selling the platform.
each sentence can look harmless alone. the accumulated effect is worse. readers do not usually write to complain that a company’s pronouns changed from “we” to “the organization.” they simply stop hearing a recognizable person behind the copy.
a brand voice review examines a draft as a match problem, not as one universal writing-quality score. a human-written paragraph can be completely off-voice. an ai-assisted paragraph can earn its place after an editor replaces the vague setup, restores customer language, and removes the polished filler that arrived with the first draft.
across the writing we have studied in voice-profile reviews, the repeat offender is vocabulary substitution. a company that normally says “help” starts saying “enable.” “show” becomes “surface.” “fix” becomes “optimize.” none of these swaps is automatically wrong. they matter when they appear together and erase the language a reader has seen in emails, product screens, and support docs.
consider two versions of the same idea:
“your renewal date is in account settings. change it before the current term ends.”
“our comprehensive platform enables customers to optimize their subscription journey.”
the second sentence may describe a real feature. it also avoids every useful noun. there is no renewal date, setting, term, or action. it could sit on a jasper landing page, a fintech brochure, or a forgotten deck from 2019.
hold your voice flags the gap for editorial judgment. it does not declare that formal language is bad or that every repeated phrase is evidence of ai. legal copy, regulated industries, and product terminology sometimes require language a brand would not choose in a founder email. the useful output is a short list of places worth reading with suspicion, not a fake verdict about writing talent.
how does hold your voice find brand voice drift in a draft?
hold your voice finds drift by looking for clusters of changes in tone, phrasing, sentence shape, vocabulary, and point of view against the reference writing you provide. one formal sentence can be deliberate. five paragraphs that suddenly read like generic category copy are a review problem.
the product review starts with a draft and examples of what “normal” sounds like. those references can be published substack posts, lifecycle emails from convertkit, help-center articles, founder notes, or a short voice guide. a 40-page brand book is rarely necessary. three strong examples and a list of phrases the company would never use can expose a surprising amount.
the checks focus on practical signals:
- tone that moves from plain and useful to promotional or stiff
- familiar customer words replaced with abstract category language
- repeated setup lines, padded transitions, and conclusions that repeat the opening
- point-of-view changes, such as a founder’s “i” becoming an impersonal corporate narrator
- sentence-length flattening, where short direct sentences disappear under evenly shaped explanations
in our analysis of voice profiles, the fastest visible drift often appears after a draft has passed through an ai first pass and one rushed human cleanup. the writer keeps the topic and facts, but loses signature transitions and sentence-length variation. by the third ai-shaped paragraph, every sentence begins to explain, qualify, or summarize. the raw notes had the useful phrase. the polished draft removed it because it looked untidy.
a support draft might begin, “your renewal date is in account settings.” later it says, “our comprehensive platform empowers customers to optimize their subscription journey.” hold your voice treats that second line as a vocabulary and specificity mismatch. “empowers” belongs on a watchlist because it carries no operational instruction. “subscription journey” hides the actual task. an editor can replace it with the setting name, the action, and the consequence.
you can check voice consistency across a draft before final approvals, when revisions are still cheap. that matters for agencies moving between client accounts, where each account has different tolerance for promotional language. it also matters for marketing teams, where a product page, paid ad, and onboarding email can be written by different people in the same week.
what most guides get wrong is treating consistency as repetition. copying the same phrases everywhere creates another failure: the brand starts sounding like it has one approved sentence. voice consistency is a pattern of choices, not a phrase bank. a good review preserves the company’s level of directness, its preferred nouns and point of view willingness to say what happened without adding a cloud of claims around it.
for examples of the warning signs, see voice drift signs in writing. the decision after a flag stays human. keep the line when the channel, audience, or legal requirement calls for it. change it when the line is merely hiding that nobody wrote a specific sentence.
can hold your voice review ai-assisted writing without pretending to know who wrote it?
hold your voice can review ai-assisted writing for patterns that make it feel generated or disconnected from a known voice, without claiming to prove who wrote it. authorship probability is a weak substitute for an editorial decision, especially when the real issue is flat language that says little.
chatgpt can help with an outline, a research summary, a first draft, or five alternate subject lines. a freelancer can send copy that resembles ai output without using ai at all. an internal writer can produce generic text on a tired thursday. the published page still carries the company name, so the editorial question remains the same: does the copy do useful work in the intended voice?
ai-shaped drafts often share a few visible habits. they open too broadly, use inflated claims where a concrete example should sit, stack transitions such as “additionally” and “in conclusion,” and give every paragraph the same calm, explanatory rhythm. they can also borrow generic emotional framing: “businesses are navigating an increasingly complex environment.” no reader learns what changed, who is affected, or what to do next.
research on ai-text detection supports caution. detector outputs can be unreliable across writing conditions, and openai retired its own text classifier in 2023 because of low accuracy. see openai’s notice at https://openai.com/index/new-ai-classifier-for-indicating-ai-written-text/. a detector score may be interesting, but it cannot tell an editor which sentence to rewrite or whether a writer acted dishonestly.
the workflow is deliberately plain. paste the draft. add reference writing or voice guidelines. review the flagged sections, and revise the weak lines, then run the review again. the second pass matters because quick fixes often introduce a new mismatch: one sentence suddenly sounds like paul graham while the rest sounds like the company’s support team.
for example, an ai draft might say, “teams need better visibility to make informed decisions.” a revision based on product reality could say, “you can see which invoices failed before finance exports the monthly report.” the first line has the familiar shape of category copy. the second names a user, a failure, and a moment in the workflow.
ai-like writing patterns are useful review signals, not evidence of authorship. hold your voice treats them that way. it looks for generic framing, phrase reuse, abstraction drift, and sentence patterns that clash with the supplied references. it does not accuse a person of using chatgpt or claim a probability score can settle the question.
that limit matters for bylines and editor handoffs. an editor needs a reason to mark a line, not a machine-generated label pasted beside it. the line should be marked because it stopped sounding like the author, stopped naming the reader’s actual problem, or started repeating a structure that the rest of the draft does not use.
who should use hold your voice before publishing?
hold your voice is useful for anyone publishing copy that has passed through multiple hands, ai tools, or fast approval cycles. the risk rises when the draft is touched by a founder, an agency writer, chatgpt, a client reviewer, and an editor with ten minutes left before the scheduled send.
individual writers use it when ai helps them get moving but starts sanding down their opinions. a writer’s voice often lives in choices that generic tools remove: the short objection after a claim, the odd but accurate customer phrase, or the willingness to leave a sentence blunt. justin welsh and alex hormozi can both be imitated at the surface level. that imitation is not a voice system. it is a recognizable cadence applied to someone else’s facts.
content leads use it to triage. when five freelancers submit articles, reading every line with equal attention is slow and usually unnecessary. a review can identify the sections where tone, wording, and structure change sharply enough to deserve close editing. it does not replace an editor who knows the subject. it gives that editor a better first pass.
agencies face a more expensive version of the problem. writers move from a healthcare client to a saas client, then from a product launch to a support update. the same writer can carry phrases across accounts without noticing. “drive results” is one of those phrases. it survives because it sounds acceptable in isolation and means almost nothing in context.
marketing teams need the review when lifecycle email, paid landing pages, social posts, sales material, and product pages all describe the same product in different language. readers notice the mismatch quickly, and the sales page promises control, then the onboarding email says simplicity. the help center says “contact support.” none of those statements may be false. together, they suggest nobody owns the language.
founders and subject-matter experts benefit when their raw notes contain the detail, but the final draft comes back sounding like category copy. their notes often include the exact operational point that a polished draft deletes. “customers kept emailing because the renewal date was buried” is stronger than a claim about improving the subscription experience.
start with the draft already waiting for approval. bring a few reference examples and a list of language to avoid. then analyze a draft against your brand voice before it goes out. the tool will not create a perfect voice guide from thin air, and it cannot decide what your company should believe. it can show where the current draft has stopped sounding like the company that published the last one.
frequently asked questions
what is hold your voice?
hold your voice is a writing review tool that helps writers and teams find brand voice drift and ai-like patterns before publishing. it checks whether a draft still sounds like the person, company, or publication behind it.
how does hold your voice detect brand voice drift?
it reviews changes in tone, vocabulary, phrasing, sentence structure, point of view, and repeated writing habits against reference writing or voice guidelines. flagged sections are prompts for editorial review, not automatic verdicts.
is hold your voice an ai detector or a brand voice checker?
it is primarily a brand voice review tool. it can flag ai-like patterns such as generic openings, padded transitions, and flat sentence rhythm, but it does not claim to prove who wrote a draft.
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.