4 ways to get more from AI writing and editing, without losing your voice

Lorra M. Brown is director of the online Master of Arts in Digital Communication program and an assistant professor at the University of North Carolina at Chapel Hill’s Hussman School of Journalism and Media.

Talking to senior communicators earlier this year surfaced a shift in how leaders are defining the job in this AI era.

A chief operating officer of a global agency network summed up the shift this way: “When AI expedites execution, the value of a communicator lives in judgment and business fluency. AI can produce content, but production is no longer the differentiator. What’s missing is context: audience nuance, organizational politics, risk.”

The distinction between producing content and adding judgment matters more now than it did just a year ago.

Here are four ways to build AI use around human authenticity, critical perspective and judgment. These are the same judgment calls I build into how I teach PR and strategic writing.

1. Turn your prompt into an editorial brief

The biggest driver of weak AI output is rarely the model. It’s the prompt.

“Write an announcement about our new brand partnership” produces exactly what you’d expect: generic, formal copy with no point of view and no recognizable voice. AI can’t supply strategic context you haven’t given it.

Before you ask a tool to draft anything, work through six things:

1. Role: Who/what is the AI persona? A media relations strategist, an executive speechwriter, a skeptical reviewer? A journalist?
2. Audience: Be specific about the stakeholders who will read it and what they already believe, fear or need.
3. Task: Define the deliverable and its job. Is it informing, persuading, reassuring, recruiting, analyzing?
4. Tone: Use concrete terms like “direct and confident” or “warm but unsentimental,” not “professional.”
5. Constraints: Name what the draft must not do. Ban tired phrases like “proud to announce” and “exciting new chapter.” Include and require substantiated claims and specific proof points.
6. Format: Specify length, structure, channel and whether you want headlines, pull quotes or message variations.

A shared brief creates consistency before the writing starts, and frees writers to spend less time repairing generic language and more time on context, nuance and purpose

2. Build a workflow, not a favorite-tool list

The question I hear the most is which AI platform is best. It’s the wrong question. The better one: what role should each tool play in your workflow?

One tool may be strongest at brainstorming and fast variations. An enterprise system usually holds voice best over long, sensitive writing. A research-oriented tool surfaces more reliable sources and citations. Your organization’s approved productivity suite is a practical place to store, review and collaborate on confidential and approved materials.

A simple sequence works for most teams:

1. Research the subject, purpose and audience, and verify the evidence.
2. Set the editorial brief and pull in approved source material.
3. Draft or generate options with whichever tool fits the assignment.
4. Use AI to challenge the draft: unsupported claims, missing stakeholder reactions, questions a skeptical reporter might ask.
5. Route it through your organization’s approval process.
6. Let a human editor make the final call.

That last step is easier to defend to legal, IT and leadership than a list of favorite apps.

3. Give AI evidence of your voice

AI can reproduce patterns. It can’t infer your organization’s full voice from a few adjectives in a prompt. “Make this sound like us” isn’t a usable instruction unless you’ve shown what “us” (the brand? a senior executive?) means.

Before your team drafts anything, load five things into whatever tool you use:

• Your current voice and brand style guide
• Three to five examples of your organization’s best writing
• Your approved messaging framework, with the proof points that need substantiation
• A “don’t say” list of words, phrases and clichés to avoid
• GEO and SEO keywords that stay true to your brand’s mission, audience and voice

This is a critical step, especially when teams are producing high volume across channels. The risk is in repeated, polished, impersonal copy that erodes what makes your organization sound like itself.

Examples transfer voice better than adjectives do. A strong executive note, a clear testimonial, a well-vetted response to a hard issue: these show the AI tool how your organization handles rhythm, specificity and point of view. This isn’t a one-time setup that lets your team turn off its judgment. It’s a foundation that makes every draft easier to write and easier to evaluate.

4. Make the human editor own the decision

Even a well-briefed draft has predictable weaknesses. Train your teams to catch all four.

• It buries the lead. It frequently opens with context before news. Ask what a journalist would headline, and lead with that.
• It flattens tension. The copy is too upbeat, even when the story includes a provocative element or vulnerability. Put the stakes back in.
• It speaks in categories, not evidence. “Significant growth” and “robust results” aren’t claims. Replace the category with the number: not significant growth, 14%. Don’t say it. Prove it.
• It sounds like no one. Clean grammar, impersonal voice. Read it aloud. If your team, a senior executive, an influencer or a spokesperson wouldn’t say it in a meeting or to a reporter, it doesn’t belong in the document.

Go a step further than editing. The highest-value use of AI right now may not be generation but an editorial challenge (see points 1 and 2): ask it what a skeptical journalist would question, which claims need evidence, what stakeholder reaction you’re missing. Used that way, AI stops being a ghostwriter and starts working like a tireless first-pass reviewer, catching problems before the meeting instead of during it.

Before anything AI-assisted publishes, run it through eight quick checks:

1. Does the lead deliver the news in the first sentence?
2. Is every vague word replaced with something concrete?
3. Does it sound like your organization, not a generic brand?
4. Would any phrase make you wince read aloud, whether it’s too safe or too provocative?
5. Is there a genuine point of view, not just information?
6. Have you restored the stakes AI tends to remove?
7. Can you defend every fact, claim and structural choice?
8. Did a human, not the model, make the final call?

Every communicator has access to increasingly capable tools now. The teams that pull ahead are the ones who create workflows that require us to judge what AI can’t: nuance, tradeoffs, ethical implications, stakeholder context and the specific evidence that makes a message credible.