Steps to use AI without losing your voice
This resource is based on 5 marketers, 5 AI workflows, zero sameness, featuring five marketing experts who demonstrate there is no single "right" way to use autonomous marketing.

Get the checklist
By the end of this checklist, you’ll have a repeatable system for using AI at every stage of content creation while keeping your distinct voice intact. It’s structured around the High-Performance Marketing Triad (Imagine, Activate, Validate) and draws on workflows from Joe Pulizzi, Dan Sanchez, Kate Bradley Chernis, Gini Dietrich, and Brandi Holder. Allow 1–2 hours for the full setup; individual phases can be done independently.
Before you start
- Define your voice in writing: Document your tone, style, formality level, and the words you use and don’t use. As Brandi Holder puts it: “I am very clear on my distinct voice. The tone, the style, the formality, the words I use and don’t use. I think that is the essential starting point.”
- Gather your existing content: Collect past writing samples, blog posts, books, or newsletters that represent your voice at its best
- Choose your AI tool: ChatGPT, Claude, or another LLM you’re comfortable with
- Define your process before you automate: Know the steps you follow manually so you can decide which ones AI should touch. As Dan Sanchez warns: “You can’t automate anything unless you define the process.”
The workflow
Phase 1: Imagine (generate smarter ideas with AI)
After this phase, you’ll have: a set of AI-assisted ideas that originated from your own thinking
- Start with your own idea first: Write down the concept, challenge, or campaign direction before opening any AI tool. Brandi Holder’s rule is “AI second, never first”—the core thinking and perspective originate from you.
- Start with your own idea first: Write down the concept, challenge, or campaign direction before opening any AI tool. Brandi Holder’s rule is “AI second, never first”—the core thinking and perspective originate from you. Use the prompt below.
- Break through blank-page syndrome with an outline request: If you’re stuck, prompt AI with whatever challenge you’re facing and ask for an outline. Gini Dietrich’s approach: “Prompt it with whatever challenge I’m facing, explain why I’m stuck, and ask for an outline.”
- Stress-test your ideas: Ask the AI to identify gaps and assumptions. Holder uses AI to “identify what is missing or what I haven’t thought of.”
- Use AI for elimination, not just generation: If the AI’s suggestions don’t fit, treat them as springboards. Kate Bradley Chernis doesn’t use AI to brainstorm directly but says: “It spat out a bunch of things. I didn’t like any of them, but it gave me another idea.”
Brainstorming prompt
I’m working on [DESCRIBE_YOUR_IDEA]. Here’s what I have so far: [PASTE_YOUR_ROUGH_CONCEPT]. Build on this — what angles, hooks, or directions haven’t I considered?
Outline prompt
I’m writing about [TOPIC] and I’m stuck because [REASON_YOU’RE_STUCK]. Give me an outline that covers [KEY_POINTS_YOU_KNOW_SHOULD_BE_INCLUDED].
Stress-test prompt
Here’s my positioning concept: [PASTE_YOUR_CONCEPT]. What assumptions am I making? What’s missing from this argument?
Phase 2: Activate (build systems that amplify your voice)
After this phase, you’ll have: at least one repeatable AI-assisted workflow for your content production
- Apply the CRIT Framework to set up AI properly: Before issuing any assignment, walk through Joe Pulizzi’s four steps outlined below
- Upload your writing samples: Feed the AI your existing content so that it learns your patterns, vocabulary, and style before producing anything
- Identify your repeatable processes: Look for content tasks with clear inputs and outputs that stay consistent regardless of audience. Dietrich’s team trained AI on their proposal and RFP response patterns—“What used to take hours and hours and hours to do has been reduced to an hour or 90 minutes, tops.”
- Build a multi-format output system: Set up a workflow that takes one raw idea and produces multiple formats. Sanchez takes any raw idea and turns it into three LinkedIn posts: a long one, a short and concise one, and a more spicy take.
- Keep AI out of nuance-dependent work: If a task depends on real-time interpretation, body language, or contextual subtlety, keep yourself in the driver’s seat. Holder notes: “Pattern recognition in my work relies heavily on what’s not said: the pauses, the body language, the shift in energy.”
Joe Pulizzi’s four steps:
- Context: Upload your past work and give the AI your world
- Role: Tell it who to be (editor, strategist, brainstorming partner)
- Interview: Instruct it to ask you clarifying questions—“Ask me no more than three questions, one at a time, to clarify what I’m trying to achieve”
- Task: Then issue the assignment
Phase 3: Validate (judge whether AI output reflects your voice)
After this phase, you’ll have: a personal quality-control process for every piece of AI-assisted content
- Read aloud or use text-to-speech: Holder uses Microsoft’s read-aloud feature: “If it sounds like something I’d actually say in a conversation, it passes. If it sounds like a LinkedIn thought leader, a press release, or a little too fluffy, I rewrite it.”
- Check for your personal quirks and patterns: Look for the specific linguistic traits that make your writing yours. Bradley Chernis checks for her onomatopoeia, made-up words, and unconventional punctuation—elements that AI can’t replicate.
- Learn your AI tells: Build a list of words and phrases that signal AI-generated content in your domain. Dietrich’s team watches for words like “receipts,” “idea bursts,” and “evidence loops”—“Anytime I see those things in my colleagues’ work, I ask them to do another edit.”
- Eliminate weak words: Review for hedging language that undercuts authority. Bradley Chernis flags words like “I think,” “I just wanted to say,” “probably,” and “need.”
- Get a second pair of eyes: Have someone else edit the final piece. Dietrich’s process: “Read and reread, then have someone else edit.”
- Judge by usefulness, not artistic merit: Sanchez’s standard is straightforward: “Is this as helpful as I meant it to be?”
- Apply AI editing with style rules: Use AI for a final polish pass with specific style guidelines applied. Dietrich’s sequence: draft it yourself, get AI feedback on what’s missing, finish the piece with that feedback, then let AI edit with AP style applied.
Quick reference
- Total time: 1–2 hours for initial setup; the validation phase becomes a 10–15 minute habit per piece
- Tools needed: ChatGPT, Claude, or another LLM; text-to-speech tool (optional); your existing content library
- Key output: A repeatable system for using AI across ideation, production, and quality control while preserving your distinct voice
Ready for the full story?
Read 5 marketers, 5 workflows, zero sameness to learn how five marketing experts use AI as tool to amplify, not replace, their voices.
Related
More data from the AI Lab.