The homogenization problem
When thousands of organizations use the same AI tools with similar prompts, the outputs start to converge. Confident, authoritative, slightly optimistic, moderately formal. It’s a perfectly fine voice. It’s just not your organization’s voice — and over time, that drift is noticeable to anyone paying attention.
Quick distinction before we start: if you've taken AI in Practice, you built a voice brief for a specific executive — capturing one person's patterns so AI could ghostwrite in their voice. This lesson is different. Here you're building a brand voice brief: a team-wide standard for your organization's voice across all AI-assisted output, not one individual's. The artifact looks similar; the purpose and audience are distinct.
AI produces text that matches the statistical average of its training data. That average tends toward a corporate-professional register: clear sentences, measured tone, balanced phrasing, minimal personality. Unless you tell it otherwise, that’s what you’ll get — every time, across every output, regardless of whether your organization’s voice is warmer, sharper, more technical, more casual, or more opinionated than average.
This isn’t a criticism of AI. It’s just how it works. The fix is straightforward: you have to give AI the information it needs to produce output that sounds like you. That information is your brand voice brief.
Key Insight
AI doesn’t drift from your brand voice because it’s failing. It drifts because it never had the brief. A well-written voice brief is the single most effective thing you can do to make AI output consistently sound like your organization.
A brand voice brief doesn’t need to be long. It needs to be specific. Generic adjectives like “professional” and “approachable” don’t help AI produce distinctive output — those words apply to thousands of organizations. What works is concrete, comparative, and example-driven.
Tone description. Describe your voice in terms of contrast: “direct but not blunt,” “warm but not casual,” “authoritative but not stiff.” Contrast forces you to be specific about where the line is.
Vocabulary to use and avoid. List five to ten words or phrases that are distinctly yours, and five to ten that your organization would never say. “We say ‘communicators,’ not ‘PR professionals.’ We say ‘earn attention,’ not ‘generate buzz.’” This is more useful than any amount of tone description.
A reference sentence. Find one sentence from your existing communications that perfectly captures how you want to sound. Give it to AI verbatim and say: “Match this register.” A good example does more work than a paragraph of instructions.
Audience framing. Who you’re writing for changes how you write. A brief that says “written for senior communications professionals who are already AI-literate” will produce different output than one that says “written for marketing managers new to AI.”
Even with a good brief, AI output will drift from your voice over time, especially across a long document or when you’re generating multiple pieces. Build a habit of reading AI output aloud before approving it. If you wouldn’t say it out loud as a member of your team, it’s not in your voice. That’s the fastest quality check for voice that exists.
Write your brand voice brief and test it against AI output. By the end of this session you’ll have a reusable voice brief you can paste into any AI prompt going forward.
Describe your organization’s tone using contrasts: “We are [X] but not [Y].” Write at least three contrast pairs. If you’re doing this for your personal professional voice rather than an organization, describe yourself the same way.
List five words or phrases you use, and five you avoid. Be specific: not “formal language” but “we say X, not Y.” If you’re stuck, pull from your existing communications and look for patterns.
Find one sentence from your existing communications — an email, a press release, a social post — that sounds exactly right. Copy it into your brief as your reference sentence.
Paste your complete brief into your AI tool, then ask it to draft a short announcement (real or hypothetical) using that voice. Read the output aloud. Mark anything that sounds off.
Refine your brief based on what AI got wrong. Save the final version as your Module 4 output. This is the document you’ll paste into any prompt where voice consistency matters.
