Hire a Brand Voice Steward Who Can Kill Copy the Model Already Wrote

Who Owns the Brand Voice When Six Channels Draft It?

Six channels shipped last Tuesday and four of them sounded like different companies: the lifecycle email polite and hedged, the paid social punchy, the help center suddenly using words nobody at the company says. A brand voice steward owns that gap. The job is the voice guide the models are conditioned on, the audit that catches drift across channels, and the call on what ships. ²

This is the old brand editor function pointed at a different object. It used to mean coaching junior copywriters until they internalized the sound. Now it means conditioning pipelines: the reference documents a model writes against, the examples that get pasted into every brief, the review step that reads for voice specifically instead of folding it into a general edit. The person is upstream of the drafting and downstream of it at the same time, and almost never in the middle writing sentences.

Drift is the failure mode to hire against, and it is quiet. No single AI-drafted email is wrong. The twentieth one is bland in a way the tenth was not, because a model regresses toward the mean of everything it has seen and the mean is not your company. A steward catches that by sampling output on a cadence and reading it against the guide, not by waiting for a stakeholder to complain that something felt off.

The scope question worth settling before you post the role: does this cover synthetic speech and voice interfaces too, or only text? Those are separable jobs. If most of the risk sits in an audio persona and how it sounds in a customer's ear, the hire you want is closer to a synthetic voice designer, which carries a different craft stack entirely.

Hire the Standards Editor Who Made Two Hundred Bylines Read as One

A newsroom standards editor has spent a career writing rules that let two hundred bylines read as one masthead. That is the artifact this role produces, and it is why the strongest candidates come from standards editing rather than from brand marketing. The four channels that sounded like different companies last Tuesday are two hundred bylines with the newsroom taken out.

The rest of the reliable pool: localization program management, senior brand copywriting with guide ownership, and technical writing behind a real docs style guide. What they share with the standards editor is having been responsible for the consistency of text they did not personally write, under deadline, with other people's names on it. A localization lead has run multilingual pipelines with a review layer since long before generative tooling arrived, and already thinks about voice as something that survives translation or does not. Liturgy, legal drafting and API documentation all produce people who are calm about the sentence they are not allowed to change.

What to listen for is narrow and checkable. A candidate who has held a voice line can name a campaign they stopped and what it cost them politically. They can describe a rule they wrote, watched fail against a model's output, and rewrote, because a guide that has never been revised has never been tested. They talk about banned constructions and sentence-length norms; adjectives like warmth and confidence are what a guide says before anyone has had to enforce it. And they say the word no in the interview without softening it. Ask what happens when the head of demand generation wants to ship something that clears the bar for correct and fails the bar for voice, then check whether the answer contains a mechanism or a hope.

Fluency with the tools will not sort them. Lightcast's read on the market is that writers stand out among traditionally non-technical occupations adopting generative tooling ³, so every serious candidate now arrives fluent. That was a signal two years ago and is a floor now.

Screen the Steward on the Drafts They Refused to Ship

Ask for the refusal log. Anyone who has genuinely held a voice line can show you drafts a pipeline produced and did not publish, the rule each one broke, and what changed in the guide afterward; the help center sentence nobody at the company would say belongs in it. That artifact is hard to fake and impossible to borrow. Published work proves an approval happened; the refusals prove someone was reading with a standard in hand.

The practice that produces this person is unglamorous. They ran one brief through a model four or five times and read the outputs against each other for where the register slid. They kept a running file of constructions the model reaches for by default, the tricolons, the balanced antitheses, the openers that announce a topic before saying anything, and turned that file into a banned list. They wrote before-and-after pairs, because a model can be shown a pair and cannot be shown a vibe. Ask what their guide says about second person, and whether it says why.

Two questions do most of the interview work, and neither is about craft. Show a piece your pipeline produced and ask the candidate to mark which sentences a person actually wrote; the honest answer is specific and slightly awkward. Then ask what their guide deliberately leaves loose, and why. Anyone who has enforced a real standard knows which rules were not worth the friction, and that answer separates a steward from a pedant.

Better than either is watching it happen. Give a 45-minute exercise: a house style guide with one internal contradiction planted in it, an AI assistant, a brief, and a claim in the source material that the guide's own rules would not let you state that confidently. What you learn is whether the candidate notices the contradiction, which way they resolve it, and whether they flag the claim or smooth it over. Content roles adjacent to this one screen well the same way, which is why an AI content strategist interview tends to converge on the same exercise design.

Find the Voice Steward Where the Style Guides Get Written

Very few people are searching under this title, so search on the work instead. The candidates you want are currently employed as senior brand copywriters, editorial directors, content design leads and localization managers, and are already doing the stewardship informally because someone had to. Their manager usually does not know the role has a name yet.

Venues that reliably exist and repay watching: the Content Marketing Institute's events and community, the Write the Docs conference and its Slack, and the content design and UX writing circuit on LinkedIn and Substack where practitioners publish teardowns of their own guides. Somebody who has published a style guide in public, including the parts that did not work, has pre-screened themselves. A conference talk beats a portfolio here for the same reason a mechanism beats a finish.

Closing them turns on one thing, and it is not money. This person wants to know whether the stop is real. Can they hold a campaign, and who can overrule them? A steward with recommendation power and no authority becomes a complaints inbox within a quarter, and the strong candidates have either lived that or heard about it from someone who did. Say plainly where the authority sits, including its limits, because the ambiguity surfaces in week three either way.

What kills the offer late: discovering the real mandate was volume with a quality label on it, a reporting line into demand generation with a monthly output target attached, or a guide that legal and brand both claim to own. Remote is the default and fighting it is expensive, because the work is documents, audits and reviews with almost nothing that needs a room. The honest exception is the first sixty days, when the steward needs time beside the teams whose output they will be stopping. Relationships bought in person early are what make a later no survivable.

What Does a Brand Voice Steward Cost in 2026?

There is no salary series for this exact title, so price it off the nearest posted category and say out loud that you are doing so. One job-board aggregator carries brand voice as its own posted category and reports an average of roughly $48 an hour across those United States postings as of 2026 ¹. That is one aggregator's own category average rather than a band for a role with stop authority, so read it as a floor.

Two adjustments push the real number up. The category mixes freelance and contract copywriting work with staff roles, and it includes people who write in the voice rather than people who govern it. Governance carries headcount-adjacent weight even without direct reports, because the job's whole value is being able to overrule a channel owner. So the proxy worth posting against is your senior brand or editorial director band, at the top of it, because that is the band already paying for the standing to return somebody else's work.

One adjustment pushes the other way, and it is worth being honest about. The 2026 AI marketing job lists describing this shape of work pair copywriting expertise with AI tool fluency ², and tool fluency is now widespread rather than scarce. Do not pay a premium for the tools. Pay for the judgment that decides what does not ship, which is the part that stayed scarce.

Structure the package around scope rather than title inflation. What candidates ask about, roughly in order: whether the guide is theirs to change, whether the audit has budget and time in it or is expected to happen in the margins, whether they sit close enough to the CMO that a stopped campaign gets a hearing, and whether there is any pathway to owning tone in product surfaces too. If the answer to the last one is yes, say so, because it converts. If it is no, say that instead. This candidate pool reads evasion for a living.

See a sample report >