Gains:
- Identifying the elements that make up the author's voice (word choice, sentence rhythm, tone) and providing a voice guide to artificial intelligence
- Ability to recognize the homogenizing, generic tone of artificial intelligence and pull the text back to its own style
- Being able to try different tone and register (formality level) demands with artificial intelligence and make the final stylistic decision as a writer
Let two writers describe the same event; The texts are completely different. The difference is not in what is said, but in how it is explained. This "how" is the writer's voice: word choice, sentence length and rhythm, humor, distance, repetition, spacing. Voice is the fingerprint that makes an author recognizable in the reader's mind. At this point, artificial intelligence (AI) can be both the most useful and the most dangerous. Useful because it speeds up tone experiments; dangerous because, left to its own devices, it reduces everyone's voice to one smooth, anonymous tone. In this unit we will cover the elements of audio, giving voice guidance to the AI, and protecting your voice.
Elements that make up sound
Sound seems like an abstract concept, but it consists of concrete elements:
Word choice (diction): Simple or fancy, casual or literary, concrete or abstract. "He died" and "he passed away" are different sounds.
Sentence rhythm: Short, punchy sentences; long, flowing sentences; Or is it a contrast of the two. Rhythm is the pulse of the text.
Tone: The text's attitude toward the reader—sincere, distant, sarcastic, compassionate, authoritative. The same information feels completely different with different tones.
Register (level of formality): The formality of a legal text and the concise language of a blog post are different registers.
Oddities: Author-specific repetitions, favorite images, broken rules. This is the most personal layer of the voice and the one that AI erases the most.
The table below shows how the same sentence changes in different voices:
item
Neutral/AI default
Personal voice sample
word
"It was very cold."
"The weather was the kind that drives people indoors."
rhythm
medium length, straight
"Short. Punchy. Then a long, winding breath."
tone
informative
Sarcastic, sincere, provocative
weirdness
None
A repeated image, a conscious inverted sentence
Giving voice guidance to AI
Telling the AI to "write like me" won't work because the AI doesn't know your voice. Instead, give a voice guide: a few sample paragraphs of yours, words you like and avoid, tone notes, rhythm preferences. AI is good at extracting patterns from examples; The more concrete material you give it, the closer the output gets to you. However, this is an approach, not a copy; The final overhaul always stays with you.
Your role: a writing assistant who adapts to my style.Below are 2 sample paragraphs that I wrote — unravel MY VOICE here:[sample paragraph 1][sample paragraph 2]My voice notes: I like short sentences; I avoid fancy adjectives; I use a sarcastic but warm tone; I do not use reinforcers such as "almost, literally, definitely". I prefer concrete images to abstract concepts.Task: Write a 150-word paragraph on: [topic].Stick to the voice above; You won't slip into your "generic" tone.
You can prepare the audio guide once and reuse it in every session. Over time, your list of "words I avoid" grows, preempting the AI's specific mistakes.
Tip: Make a blacklist of “common” AI words that give you away: “sort of,” “sort of,” “nowadays,” “in every aspect of our lives,” “plays an important role.” Add these to the prompt as "do not use"; your text instantly becomes less “AI smelly”.
AI's tendency to homogenize
Since AI extracts averages from millions of texts, it naturally gravitates towards average voice. This tone is sometimes called "casual fluency": a language whose grammar is perfect, whose flow is smooth, but which takes no risks, never surprises, belongs to no one. Its symptoms are: overly balanced sentences, clichéd transitional expressions ("on the other hand", "in conclusion"), closing each paragraph with a generalization, deadpan neatness. Recognizing this tone is the first step to escaping it. While reading a text, "can anyone write this?" ask; If the answer is yes, it's time to bring your voice back.
Tone and register experiments
One of the most legitimate uses of AI is to see and make choices about the same text in different tones. You can produce a promotional text in "serious", "humorous" and "emotional" tones and see which one suits your brand. Here, AI is a test bench; The final stylistic decision is yours.
Task: Rewrite the following paragraph in 3 different tones: 1) formal and reassuring, 2) warm and friendly, 3) short and striking. RETAIN the meaning; just change the tone. Add a single sentence note under each version, "what context does this tone fit in?"
This template makes it easier for you to decide, but it doesn't make the decision for you. You'll see three options, choose one or blend it, then add your own finishing touch.
"Calibrating" the sound: feedback loop
The audio guide is not a one-off document, but a living tool that is being developed. Establishing a small feedback loop when working with AI brings the output a little closer to you each time. The loop works like this: AI produces a paragraph; you manually correct it to your own voice; Then you tell the AI, "here's what I fixed — look at what I changed and put those preferences into the rule." The AI extracts patterns from your corrections (“shortens long adjective chains, deletes reinforcers, ends short sentences”) and you add these rules to your audio guide. After a few cycles, the guidance becomes so sharp that even the initial output of the AI comes noticeably closer to your voice.
Here is the paragraph that AI wrote and my hand-edited version of it:[AI version][my version]Task: Compare two texts. What changes did I make? Put them in a 5-item "voice rule" list so that next time you can write according to these rules from the beginning. Don't add comments, just remove the rules.
Caution: This cycle is powerful, but also dangerous: if you "teach" your voice entirely to the AI and start making it write everything, your own muscles will atrophy. Use the guide as a catalyst, not as a substitute imitator.
three mini cases
Case 1 — Voice recovery. One columnist noticed that the drafts he wrote with AI "didn't look like him." He gave two of his old articles as audio guides and added a blacklist. The new drafts came noticeably closer to it; Reader comments said "your old voice again".
Case 2 — Tone experiment. A brand team couldn't agree on the tone of a campaign copy. They had the AI produce the same text in three tones; They saw that the "warm and friendly" version suited the target audience and added their own touches to it. The decision became clear within minutes.
Case 3 — Homogeneity trap. One author had ten blog posts written entirely by AI; they all had the same smooth, anonymous tone, and reader engagement dropped. He realized that writing had become such that "anyone could write"; He reworked the texts with the audio guide and personal observations, and the connection was re-established.
Weak prompt / Strong prompt
Weak prompt:
Write this in my style and in beautiful language.
The AI thinks “beautiful language” is its average tone; The result is anonymous.
Powerful prompt:
Below are 2 examples from my writing, unravel my voice here: [examples]Words I avoid: almost, nowadays, play an important role, sort of.I like: short sentence, concrete image, sarcastic-warm tone.Task: Write 120 words about [topic]. DO NOT shift to general/average tone; end each sentence with "would I write that?" pass the test.
Difference: concrete example, blacklist, sound test — the output gets closer to you.
Common mistakes
- Saying "Write like me" means not giving examples. AI doesn't know your voice; shifts to the mean without sample.
- Mistaking the commonplace fluency of AI as "quality". Just because it's smooth doesn't mean it belongs to you.
- Not keeping a blacklist. If you do not prevent AI's unique clichés, it will repeat itself in every text.
- Leaving the tone decision to the AI. AI provides choice; You decide which tone suits you.
- Skipping the final overhaul. Even the audio guide is an output approach; The final touch comes from the author.
In summary
Voice is the fingerprint that makes an author recognizable in the reader's mind, and it is inalienable. AI is a powerful machine that accelerates tone experimentation, but left to its own devices, it reduces everyone to a single, anonymous tone. Solution: Giving the AI a voice guide of concrete examples and a blacklist, outputting “would I write this?” test it and always make the final touch yourself. AI tries tone; You carry the voice.
Application task
Choose two paragraphs of your own writing and make a sound guide out of them: words you like, ones you avoid, notes of rhythm and tone. Also make a personal blacklist. With the audio guide template in this unit, ask the AI for a paragraph and output the question “Would I write this?” pass the test. Then, with the tone trial template, produce the same text in three tones and choose one with justification.
checklist
- [ ] Have I given the AI an audio guide of concrete examples?
- [ ] Have I made a personal blacklist of "words to avoid"?
- [ ] Change the output to "Would I write this?" Did I pass the test?
- [ ] Have I noticed and corrected common AI fluency symptoms?
- [ ] Did I see the options and make the tone decision myself?
- [ ] Did I make the final touch with my own hand?