A SaaS company in the workforce management space hired its tenth content writer in three years. Each of the previous nine had produced solid work, but each had also brought their own voice: the second writer’s casual aside, the fifth writer’s love of em dashes, the seventh writer’s tendency toward corporate buzzwords. By the time the tenth writer joined, the company’s blog read as written by ten different people. The brand had no recognizable voice. It had ten voices in a trench coat.
This is what happens to companies that scale content without a real voice guide. The work gets published, the volume looks impressive, and the brand identity gets eroded a piece at a time. By the time the leadership team notices, the rebuild requires either rewriting the archive or accepting that the brand has lost any distinct way of speaking.
This piece is about the work that prevents that. What a real voice guide contains, how to build one, and how to use it across human writers, AI tools, and the multi-channel reality of modern publishing.
What a voice guide actually does
A voice guide is the document that tells a writer, an editor, an AI assistant, or anyone else producing content for the brand how that content should sound. Not what to write about, not how to format it, not what corrections to apply at the end. How it should read.
The decisions a voice guide answers include: how formal or casual the writing should be, how direct or hedging, how warm or businesslike, how technical or accessible. It also answers questions about specific stylistic patterns: do we use the first person, second person, or third person? Do we use contractions? How long are sentences? How long are paragraphs? Do we use rhetorical questions? Do we use humor, and if so, what kind?
The voice guide also names the things the brand does not do. The phrases it would not use, the genres of post it would not publish, the registers it would not adopt. The constraints are as important as the affirmative description because they prevent voice drift.
Without this document, every writer ends up making these decisions independently, every piece gets reviewed against an implicit standard that lives in the editor’s head, and every new hire takes months to absorb a voice that nobody can articulate. With it, the team produces consistent work and new contributors ramp in weeks rather than quarters.
The structure of a voice guide that works
A useful voice guide tends to have the following sections.
A voice statement of two to four sentences that captures the essence. Not “we are bold and innovative.” Something like “We sound like a senior practitioner explaining their work to a peer over coffee. We are direct and specific, willing to take positions, and uncomfortable with vagueness. We trust the reader to handle complexity.”
A description of the audience the voice is calibrated for. Who they are, what they know, what they expect from credible writing in this space. The voice exists in relation to an audience and naming the audience clarifies the calibration.
A set of voice principles, four to seven of them, each described with a paragraph. Examples of principles a real voice guide might use: “Be specific over general.” “Use plain words over jargon when both work.” “Take a position rather than hedging.” “Vary sentence rhythm, do not chant.” “Trust the reader to handle complexity.” Each principle gets unpacked with examples.
A do-and-don’t section with concrete examples. “We say ‘this approach failed because the customer hated the onboarding flow.’ We do not say ‘this approach faced challenges with the customer experience.’” Twenty to thirty paired examples like this give writers a pattern library to ground against.
A banned-words list. Words and phrases the brand does not use. Marketing buzzwords like “leverage,” “unlock,” “robust,” “seamless,” “cutting-edge,” and “revolutionary.” Throat-clearing phrases like “It is worth noting that…” and “In today’s digital age…” AI-generated tells like “navigate the landscape” and “dive deep into.” This list does work because it is the most actionable part of the document.
A register section that maps the core voice to specific channels. Slightly more formal in help-center docs, slightly more conversational in social, etc. With examples for each.
A positioning section that names what the brand specifically does not sound like. The competitors whose voices we would not want to be confused with, with examples. This sharpens the affirmative description.
A grounding section with three to five short pieces of writing that exemplify the voice well. These reference pieces give writers something to read and absorb, which moves voice faster than abstract descriptions can.
How to actually develop the voice
Most companies try to develop a voice by sitting in a conference room and brainstorming adjectives. This produces voice guides that say “we are friendly but professional, bold but accessible, technical but approachable.” None of these produce identifiable voice in actual writing.
The development process that works is more empirical. It starts with the writing the brand has already produced, identifies the patterns in the writing the team likes, codifies those patterns, and tests them against the writing the team does not like.
Specifically. Pull twenty pieces of past writing the team agrees represents the brand at its best. These can be blog posts, marketing emails, help-center articles, executive LinkedIn posts, anything published. Read them carefully and note the patterns. Sentence length distribution. Use of first vs. third person. Use of rhetorical structure. Use of specific kinds of evidence (case studies, data, examples, quotes). Common opening moves. Common closing moves.
Then pull twenty pieces of writing the team agrees does not represent the brand. Sometimes these are pieces from competitors. Sometimes they are early drafts the brand rejected. Sometimes they are pieces the team felt drift in but published anyway. Identify the patterns there too.
The voice guide then articulates the patterns from the first list as principles and the patterns from the second list as banned moves. This produces a voice description grounded in what the brand actually sounds like at its best, rather than aspirational adjectives that no one knows how to act on.
The role of examples and counter-examples
The single most useful pages in any voice guide are the example pages. A principle like “be specific” is less useful than a paired example showing what specificity sounds like in this brand’s voice versus what generic writing sounds like.
Bad example: “We worked with our customers to improve their workflows.” Good example: “We sat with three customer success teams while they worked through tickets and rebuilt the workflow editor based on what we saw.”
The first sentence is the kind of corporate phrasing that has zero personality. The second sentence is specific, grounded in observation, and shows the brand thinks about its work as practitioners think about their work. Showing this paired example does more to teach voice than any number of principle statements.
A working voice guide has dozens of these paired examples. They are tedious to write but they are the most-referenced pages in the document.
How to use the voice guide with AI tools
AI assistants have entered most content workflows. They draft, they outline, they edit, they suggest. The voice guide has to work with them or against them.
The pattern that works: include the voice guide content in the system prompt or instruction context whenever AI tools are doing meaningful writing work. Models follow concrete pattern examples better than they follow abstract descriptions. A short instruction that says “use this brand voice” without examples will produce generic AI prose. An instruction that includes ten paired examples and the banned-words list will produce output that reads much closer to the brand.
Equally important: train the human reviewers to recognize when the AI has slipped voice. Common AI patterns to watch for include the throat-clearing opener (“In today’s rapidly evolving landscape…”), formulaic transitions (“Moreover,” “Furthermore,” “Additionally”), three-item lists used relentlessly, the closing “in conclusion” paragraph that summarizes what was just said. These are predictable enough to flag mechanically and edit out.
How to roll out the voice guide
The voice guide that gets written and never read does not solve anything. The rollout matters.
Start with an internal training session for everyone who writes for the brand: marketing, content, executives who post on LinkedIn, customer success who sends emails, sales who writes proposals. Walk through the principles, the examples, the banned words. Make sure each person understands the audience the voice is calibrated for.
Apply the voice guide to a specific recent piece of writing that did not quite hit the voice. Show the original, then a revised version. This concrete demonstration moves understanding faster than any explanation.
Build the voice guide into the editorial review checklist. Editors should be checking specifically for voice consistency, not just grammar and structure. New writers should get voice-focused feedback in their first three to five pieces.
Revisit the voice guide annually. Brands evolve, audiences evolve, the platforms where the writing lives evolve. The guide that worked in 2024 may need updating for 2026.
When voice guides do more harm than good
A voice guide can be too prescriptive. If every sentence has to follow a template, the writing becomes sterile. The best voice guides describe the patterns clearly enough to ensure consistency without choking out the individual writer’s contribution.
The signal that a voice guide is too prescriptive is when the writing starts feeling produced rather than written. When every piece sounds like the same person wrote it, even though five people did, the brand has gained consistency but lost humanity. The fix is usually to loosen specific rules and trust writers more on rhythm, structure, and lexical choice while keeping the principles and banned words tight.
A voice guide that produces consistent, recognizable, human-sounding writing across multiple writers is the goal. Achieving it takes a real document, real examples, real onboarding, and real editorial discipline. The companies that put this work in produce content that reads as a coherent brand. The ones that do not produce ten voices in a trench coat.