L1
·
Quiz
·
Lab
L2
·
Quiz
·
Lab
L3
·
Quiz
·
Lab
L4
·
Quiz
·
Lab
Module Test
Module 3 · Lesson 1

What Voice Actually Is

Before you can protect your voice, you have to know what it's made of.
What are the measurable, teachable components of a writer's distinctive voice — and why does AI erase them by default?

In September 2023, the Authors Guild released a survey of 1,174 professional authors. Among respondents who had used AI writing tools, the most common complaint was not factual error or copyright — it was homogenization. Authors described submitting AI-assisted drafts to editors who returned them with notes like "this doesn't sound like you" even when the factual content was accurate and the structure was sound. The problem wasn't what the AI said. It was how the AI said it — or rather, how it erased how they said it.

The Four Layers of Voice

Researchers in computational stylistics have spent decades trying to fingerprint authorship. The field — called stylometry — has produced a working taxonomy of what makes one writer sound different from another. These are not abstract qualities. They are measurable patterns in text.

The four layers that consistently distinguish one author from another are: syntactic rhythm (how sentences are built and how long they run), lexical selection (which words a writer chooses when many would work), tonal register (the emotional temperature and social distance a writer maintains with the reader), and structural logic (how a writer organizes ideas — chronologically, spatially, associatively, contrastively).

Patrick Juola at Duquesne University demonstrated in 2013 — in the widely publicized unmasking of J.K. Rowling as Robert Galbraith — that these four layers are so consistent they can identify an author from a sample of unknown text with high accuracy. The irony is that the same patterns AI tends to flatten are exactly the ones that make authorship identifiable.

Why AI Defaults to the Middle

Large language models are trained to predict the most probable next token given a context. Probability, by definition, favors the common. The most probable phrasing, sentence length, and word choice for any given semantic intention is the phrasing that appears most frequently in training data. That phrasing tends to be average — it is the midpoint of thousands of different writers' choices, none of them yours.

This is not a flaw in the technical sense. It is the model doing exactly what it was optimized to do. The problem arises when writers paste their own rough drafts in, ask the model to "improve" or "clean up" the prose, and receive back text that is cleaner in a grammatical sense but stripped of the idiosyncratic choices that made the original prose theirs.

In 2022, researchers at the University of Copenhagen analyzed AI-assisted student essays and found that even minimal AI intervention — as little as one round of "polish this paragraph" prompting — measurably reduced stylistic variance. The essays became more grammatically uniform and more stylistically anonymous simultaneously.

Why This Matters for Professional Writers

Editors, agents, and readers recognize voice before they can articulate it. When AI homogenizes your prose, it doesn't just change how you sound — it removes the signal that makes your work identifiably yours in a crowded market. Voice is competitive advantage. Losing it to convenience is a real professional risk.

The Measurable Markers of Your Voice

Sentence length distribution: Writers vary in how long their sentences run and how much variance they allow. Cormac McCarthy famously refuses punctuation that would create syntactic subordination. David Foster Wallace employed embedded subordinate clauses that ran for hundreds of words. Both are extremes — but every writer has a signature distribution that readers feel even when they can't name it.

Hedge frequency: Some writers hedge constantly ("perhaps," "it seems," "one might argue"). Others make declarative assertions. The ratio is a voice marker. AI systems tend toward moderate hedging because it reduces the probability of confident error — but this default may conflict with your register entirely.

Concrete-to-abstract ratio: Writers differ sharply in how much concrete sensory detail they use versus abstract claims. Joan Didion's journalism is dense with specific named objects. Susan Sontag's essays operate at a high level of abstraction. Neither is objectively better. Both are voices. AI defaults toward a moderate ratio that satisfies neither extreme.

Stylometry The quantitative study of literary style, using statistical analysis of word frequency, sentence length, and syntactic patterns to attribute authorship or detect voice drift.
Voice Drift The gradual erosion of a writer's distinctive style markers through repeated AI-assisted editing, resulting in prose that is grammatically correct but stylistically anonymous.
Lexical Selection The specific words a writer chooses when multiple synonyms would convey the same meaning — one of the strongest individual voice markers in stylometric analysis.
Core Principle

Your voice is not a feeling or an aspiration — it is a set of reproducible patterns in your writing. Once you can name those patterns, you can protect them. The writer who cannot describe their own voice cannot instruct an AI to preserve it.

Lesson 1 Quiz — What Voice Actually Is

Five questions · Select the best answer for each.
1. Patrick Juola's 2013 stylometric analysis famously unmasked which author writing under a pseudonym?
Correct. Patrick Juola at Duquesne University used stylometric analysis of sentence length, word frequency, and syntactic patterns to identify Rowling as the author of The Cuckoo's Calling in 2013.
Not quite. Juola's 2013 analysis focused on J.K. Rowling, who had published The Cuckoo's Calling under the name Robert Galbraith. The analysis examined the novel's syntactic and lexical patterns.
2. According to the Authors Guild's 2023 survey, what was the most common complaint from authors who had used AI writing tools?
Correct. The 1,174-author survey found that homogenization — prose losing its distinctive character — was the most frequently cited problem, even above factual errors or copyright concerns.
Not quite. While those are real concerns, the 2023 Authors Guild survey of 1,174 professional authors found homogenization — the loss of individual voice — was the top complaint from AI tool users.
3. Why do large language models default toward stylistically average prose?
Correct. LLMs are trained to predict the most probable next token. The most probable phrasing for any semantic intention is the most common phrasing across training data — which is, by definition, average across many different writers' choices.
Not quite. The core reason is probabilistic: LLMs predict the most likely next token, and the most likely phrasing in any context is the average of thousands of writers' choices — not any individual writer's distinctive selection.
4. Which of the following is NOT one of the four measurable voice layers identified in the lesson?
Correct. The four measurable voice layers are syntactic rhythm, lexical selection, tonal register, and structural logic. "Emotional authenticity" is a subjective quality, not a measurable textual pattern that stylometry can reliably identify.
Not quite. The four measurable layers described in this lesson are syntactic rhythm, lexical selection, tonal register, and structural logic. "Emotional authenticity" is not among them — it is a subjective quality, not a measurable textual marker.
5. What did University of Copenhagen researchers find in their 2022 analysis of AI-assisted student essays?
Correct. The 2022 Copenhagen research found that even a single round of "polish this paragraph" AI prompting measurably reduced stylistic variance — essays became simultaneously more grammatically uniform and more stylistically anonymous.
Not quite. The 2022 University of Copenhagen study found the opposite: even minimal AI intervention — one round of polishing — measurably reduced stylistic variance, making essays more grammatically uniform and more stylistically anonymous at the same time.

Lab 1 — Mapping Your Voice Markers

Practice exercise · Identify and articulate your own stylistic patterns with AI assistance.

What You'll Do

In this lab you'll work with an AI assistant to identify and document your own voice markers — the four measurable layers from Lesson 1. Bring a short sample of your own writing (150–300 words from any past work) and use the prompts below to analyze it.

If you don't have a writing sample handy, describe your usual writing style in response to the opening question. The assistant will help you excavate your patterns.

Start by typing: "Here is a sample of my writing: [paste your text]" — or — "I don't have a sample, but I usually write [describe your style]."
Voice Mapping Assistant
Lab 1
Welcome to the Voice Mapping Lab. I'm here to help you identify the specific, measurable patterns that make your writing yours — what stylometrists call your voice markers.

Paste a short sample of your writing (150–300 words works well), or describe how you typically write. I'll help you map your syntactic rhythm, lexical preferences, tonal register, and structural tendencies. Ready when you are.
Module 3 · Lesson 2

The Voice Brief

A written specification for your style — the single most effective tool for AI-assisted writing that sounds like you.
How do professional writers translate their voice into a document that AI can actually use as a constraint?

In 2022, a coalition of journalists at The Atlantic began experimenting with what they internally called "style memos" — detailed written descriptions of each writer's characteristic patterns, originally created for new editors joining the staff. By 2023, several of these same writers had begun using adapted versions of those memos as system prompts when working with AI research and drafting tools. The memos that worked weren't vague ("I write in a literary style") — they were specific: "My sentences average 18–24 words. I use em-dashes rather than parentheses for asides. I avoid the word 'utilize.' I open paragraphs with a concrete claim, not a question."

The approach spread. By 2024, "voice briefs" had become a documented practice discussed at journalism conferences including the Online News Association annual meeting, where multiple panelists cited them as essential infrastructure for AI-assisted editorial work.

What a Voice Brief Contains

A voice brief is a written document — typically 300–600 words — that translates your stylistic patterns into explicit, actionable specifications. It is not a style guide for others to imitate you. It is an instruction set for AI tools to constraint-generate within your parameters.

The most effective voice briefs contain six elements: a sentence rhythm specification, a vocabulary profile, a punctuation and formatting signature, a tone calibration, a list of explicit prohibitions (words or constructions you never use), and a structural tendency statement.

Sentence rhythm specification: "My sentences run 12–30 words. I use short sentences (under 8 words) for emphasis at most once per paragraph. I avoid passive voice except when the agent is genuinely unknown or unimportant."

Vocabulary profile: "I prefer Anglo-Saxon root words over Latinate equivalents (use vs. utilize, start vs. commence, help vs. facilitate). I draw on specific concrete nouns rather than category nouns (not 'a vehicle' but 'a 1987 Dodge pickup')."

Explicit prohibitions: These are your never-list. Every writer has words they avoid on principle or through instinct. Making the list explicit gives AI a hard constraint rather than a soft preference.

Tone Calibration: The Hardest Part

Tone is the hardest voice element to specify because writers often describe it in vague emotional terms that AI cannot operationalize. "Warm but authoritative" means nothing to a language model. What does mean something is a behavioral specification: "I address the reader as a peer, not a student. I use 'you' and 'we' rather than the impersonal 'one.' I acknowledge uncertainty in first person ('I don't know') rather than hedging with adverbs ('it is perhaps the case')."

In 2023, the content marketing firm Animalz published an internal analysis of 200 AI-assisted articles. They found that articles generated with specific tonal behavioral specifications (not emotional descriptors) were rated by editors as significantly closer to the target writer's voice than articles generated with emotional descriptors alone. The difference was not minor — articles with behavioral specs were rated "on-voice" at 71% versus 34% for emotional descriptors.

Common Voice Brief Failure Modes

Too abstract: "Literary, thoughtful, engaging" — these are aspirations, not constraints. AI has no mechanism to operationalize them.

Too generic: "Clear and concise" describes every decent writer. Your brief must describe what makes you different from other clear and concise writers.

Missing prohibitions: The never-list is often more useful than the always-list. Knowing what you never do eliminates large swaths of AI default behavior.

Iterating Your Brief

A voice brief is not written once. It improves through use. Each time an AI tool produces output that doesn't sound like you, the miss is data. Ask yourself: which element of my brief failed to prevent this? Add a more specific constraint. Writers who iterate their voice briefs over months develop documents that are genuinely precise — specific enough that even editors who don't know the author can reliably identify when AI-generated text falls outside the parameters.

The practical workflow recommended by the Online News Association 2024 panel: keep your voice brief in a separate document, paste it as the opening of every AI system prompt, and maintain a "misses log" where you record AI output failures and the brief addition that would have prevented them.

Voice Brief A 300–600 word written specification of a writer's stylistic patterns, formatted as behavioral constraints and prohibitions for use as an AI system prompt prefix.
Behavioral Specification A tone description framed as observable textual behavior ("I use 'you' and 'we'") rather than an emotional descriptor ("warm and approachable"), making it actionable for AI constraint-generation.
Never-List The explicit prohibition section of a voice brief — words, constructions, and rhetorical moves the writer categorically avoids, providing hard constraints that override AI defaults.
Core Principle

The voice brief converts your intuitive stylistic instincts into explicit constraints. The more specific and behavioral your brief, the less room AI has to default to average. Vagueness in your brief is an invitation for homogenization.

Lesson 2 Quiz — The Voice Brief

Five questions · Select the best answer for each.
1. According to the Animalz 2023 internal analysis, what percentage of AI-assisted articles with behavioral tonal specifications were rated "on-voice" by editors?
Correct. The Animalz analysis found 71% of articles generated with specific behavioral tonal specifications were rated on-voice, compared to only 34% for articles generated with emotional descriptors alone.
Not quite. The Animalz 2023 analysis found that 71% of AI-assisted articles with behavioral specifications were rated on-voice by editors — compared to 34% for those with only emotional descriptors.
2. What is the ideal word count range for an effective voice brief?
Correct. The lesson specifies 300–600 words as the effective range for a voice brief — long enough to capture meaningful specificity across all six elements, but short enough to function as a system prompt prefix in AI tools.
Not quite. According to the lesson, an effective voice brief runs 300–600 words — specific enough to constrain AI behavior meaningfully, but concise enough for practical use as a system prompt prefix.
3. Why is "warm but authoritative" an ineffective tonal descriptor for a voice brief?
Correct. Emotional descriptors like "warm but authoritative" give AI no behavioral mechanism to operationalize. What works is behavioral specification: "I address the reader as a peer, use 'you' and 'we,' and acknowledge uncertainty in first person."
Not quite. The problem is that emotional descriptors like "warm but authoritative" cannot be translated into specific textual behavior by a language model. Behavioral specifications ("I use 'you' and 'we' rather than 'one'") give AI something it can actually apply.
4. What practice did the Online News Association 2024 panel recommend for improving voice briefs over time?
Correct. The ONA 2024 panel recommended keeping the voice brief in a separate document, pasting it into every AI system prompt, and maintaining a misses log — recording AI failures and the specific brief addition that would have prevented each one.
Not quite. The Online News Association 2024 panelists recommended maintaining a "misses log" — recording instances where AI output failed to match the writer's voice — and adding specific constraints to the brief that would have prevented each failure.
5. Which of these is an example of a well-formed voice brief element?
Correct. This is a well-formed voice brief element because it contains specific, behavioral constraints — an explicit prohibition, a preference, and a structural never. These give AI hard constraints it can actually apply, unlike vague aesthetic descriptors.
Not quite. Well-formed voice brief elements are specific and behavioral, containing concrete constraints and explicit prohibitions. "I avoid 'utilize.' I never open a paragraph with a rhetorical question" is the model — it gives AI something to actually enforce.

Lab 2 — Building Your Voice Brief

Practice exercise · Draft all six elements of your personal voice brief with guided AI assistance.

What You'll Do

In this lab you'll draft each of the six voice brief elements from Lesson 2: sentence rhythm, vocabulary profile, punctuation signature, tone calibration, explicit prohibitions (never-list), and structural tendency. The assistant will prompt you through each one.

By the end of this lab you'll have a working voice brief you can paste into any AI system prompt.

Start by typing: "Let's build my voice brief. Begin with sentence rhythm." — or ask the assistant to start wherever you'd like.
Voice Brief Builder
Lab 2
Welcome to the Voice Brief Builder. We'll work through all six elements of your voice brief: sentence rhythm, vocabulary profile, punctuation signature, tone calibration, explicit prohibitions, and structural tendency.

Each element should be specific and behavioral — not "I write clearly" but "my sentences average 15–22 words and I use a sentence under 6 words for emphasis at most once per paragraph."

Ready to start? Tell me which element you'd like to work on first, or type "Let's begin" and we'll go through them in order.
Module 3 · Lesson 3

Prompting for Voice Fidelity

The specific prompt structures that keep AI inside your stylistic parameters — and the common prompts that guarantee it will leave them.
What is the difference between a prompt that unlocks AI capability and a prompt that erases your voice?

In 2023, The Guardian published an internal experiment conducted by their features desk. Three staff journalists wrote original drafts of reported pieces and then asked AI tools to "improve," "tighten," and "polish" them using those exact words. The resulting AI-revised drafts were shown blind to five senior editors, who were asked to rate how closely each felt to the originals. All fifteen revised pieces scored below 40% voice fidelity on the editors' ratings. The editors noted a consistent pattern: the AI had correctly identified "weak" sentences by standard readability metrics and replaced them — but those "weak" sentences were often the most voice-distinctive moments in each piece. The hedges, the asides, the unexpected word choices — all removed as inefficiencies.

The Five Dangerous Prompts

Certain prompts consistently produce voice erasure regardless of which AI tool is used. Understanding why they fail is as important as knowing what to use instead.

"Improve this." This is the most dangerous prompt in a writer's toolkit. "Improve" has no specification — the AI interprets it through its own default quality metrics, which favor grammatical uniformity, sentence variety according to statistical norms, and Latinate vocabulary that scores well on readability formulas. Your idiosyncrasies are, by those metrics, flaws to be corrected.

"Polish this." Similar problem. "Polish" signals the removal of rough edges — and rough edges are often where voice lives. Cormac McCarthy's "rough edges" are what make him McCarthy.

"Make this clearer." Clarity is a style judgment. What you find clear may be deliberate density that serves your argument. What AI finds clear is the most statistically average phrasing.

"Rewrite this in a more professional tone." "Professional" is AI's strongest homogenizer. It activates defaults toward corporate neutral — Latinate vocabulary, passive voice, hedge-heavy constructions.

"Fix the grammar." This sounds safe but isn't. AI grammar correction includes "correcting" stylistic choices that happen to deviate from statistical norms — sentence fragments used for effect, comma splices for rhythmic flow, unconventional punctuation that is part of your signature.

The Five Voice-Preserving Prompt Structures

Structure 1 — The Preservation Mandate: Begin with an explicit instruction before any task. "Do not change word choice, sentence length, or tone. Preserve all stylistic choices. Your only task is [specific task]." This creates a hard constraint before the task is specified.

Structure 2 — The Boundary Description: Tell AI what the text is, not just what you want done to it. "This is a piece of narrative nonfiction in a voice that is deliberately casual, uses sentence fragments for emphasis, and avoids academic hedging. Within those constraints, [task]."

Structure 3 — The Specific Intervention Request: Replace open-ended improvement prompts with targeted requests. Instead of "improve this paragraph," use "in this paragraph, find any unclear antecedents and suggest specific pronoun fixes, without changing anything else."

Structure 4 — The Output Format Constraint: Ask for tracked changes or side-by-side comparison rather than a clean rewrite. "Show me what you would change and why, without rewriting. Present each proposed change as: [original phrase] → [proposed phrase] because [reason]." This keeps you in editorial control.

Structure 5 — The Voice Check Request: After AI generates any text, run a voice check pass. "I've provided my voice brief above. Review the text you just produced and identify every word, phrase, or sentence that falls outside my stated parameters. List them with the specific parameter they violate."

The Core Principle of Voice-Preserving Prompting

Every voice-preserving prompt structure has one thing in common: it specifies what AI should NOT change before it specifies what AI should do. The preservation mandate comes first. The task comes second. Reversing this order — task first, constraints second — consistently produces voice drift, because AI begins generating from its defaults before the constraints are registered.

The Research Behind Constraint-First Prompting

In 2024, researchers at Stanford's Human-Centered AI group published findings on prompt structure and stylistic consistency. They found that placing style constraints at the opening of a prompt (before the task specification) produced measurably higher stylistic fidelity to reference samples than placing the same constraints at the end of the prompt — even when the constraints were identical in content. The difference in constraint position alone produced a 23% increase in stylistic similarity to reference text, as measured by stylometric tools.

The explanation is structural: AI generates left-to-right, building context as it reads the prompt. Constraints encountered early become part of the generation context from the first token. Constraints encountered late are incorporated into a generation that has already begun taking shape under default assumptions.

Preservation Mandate An explicit instruction placed at the opening of a prompt — before any task specification — that prohibits AI from altering stylistic elements of the source text.
Voice Check Pass A second AI interaction, after initial generation, in which the model is asked to audit its own output against the writer's voice brief and identify deviations from stated parameters.
Constraint-First Prompting A prompt architecture in which stylistic constraints and preservation mandates appear before the task specification, shown by Stanford HCI research (2024) to produce 23% higher stylistic fidelity.
Core Principle

The words "improve," "polish," and "clean up" activate AI's default quality metrics — which are designed to produce average, not distinctive, prose. Replace every open-ended improvement prompt with a specific, bounded intervention request that begins with what must not change.

Lesson 3 Quiz — Prompting for Voice Fidelity

Five questions · Select the best answer for each.
1. In The Guardian's 2023 internal experiment, what consistent pattern did senior editors identify in AI-revised drafts?
Correct. The Guardian editors found that AI had correctly identified "weak" sentences by standard readability metrics and replaced them — but those were often the most voice-distinctive moments: the hedges, asides, and unexpected word choices that made each piece recognizably that writer's work.
Not quite. The Guardian's editors found that AI had systematically removed the most voice-distinctive elements — hedges, asides, unexpected word choices — treating them as "inefficiencies" to be corrected by standard readability metrics. All 15 revised pieces scored below 40% voice fidelity.
2. Why is "make this clearer" a dangerous prompt for voice preservation?
Correct. "Clarity" is a style judgment that AI resolves through its own defaults — favoring the most statistically common phrasing. Deliberate complexity, density, or unconventional construction that serves the writer's argument gets simplified as if it were a mistake.
Not quite. The problem is that AI interprets "clarity" through its own statistical defaults — what it finds clear is average phrasing, not necessarily what the writer finds clear. Deliberate stylistic density or complexity gets erased as if it were an error.
3. What did Stanford's 2024 Human-Centered AI research find about the position of style constraints within a prompt?
Correct. Stanford HCI 2024 found that placing style constraints before the task specification — rather than after — produced a 23% increase in stylistic similarity to reference text, even when the constraint content was identical. The early position shapes generation from the first token.
Not quite. The Stanford HCI 2024 research found that constraints placed at the beginning of prompts — before the task — produced 23% higher stylistic fidelity than the same constraints placed at the end. AI generates left-to-right, so early constraints become part of the generation context from the start.
4. What is a Voice Check Pass?
Correct. A Voice Check Pass is a second prompt after initial generation, asking AI to review what it just produced against the writer's voice brief and identify every element that falls outside the stated parameters — maintaining editorial control without requiring a full human re-read.
Not quite. A Voice Check Pass is a second AI interaction, following initial generation, in which the AI is asked to audit its own output against the writer's voice brief and list every word, phrase, or sentence that violates the specified parameters.
5. Which prompt structure best exemplifies the Preservation Mandate approach?
Correct. This prompt opens with explicit prohibitions — the preservation mandate — before specifying the task. It defines what must not change before saying what should happen. This is the defining feature of a well-formed Preservation Mandate structure.
Not quite. A Preservation Mandate places explicit prohibitions first, before any task specification. "Do not change word choice, sentence length, or tone. Your only task is to find and fix unclear pronoun antecedents" is the model — constraint first, task second.

Lab 3 — Prompting for Voice Fidelity

Practice exercise · Rewrite dangerous prompts into voice-preserving structures and test them.

What You'll Do

In this lab you'll practice converting the five dangerous prompt patterns from Lesson 3 into voice-preserving prompt structures. Bring a short piece of your own writing (any length) and practice applying the Preservation Mandate, Specific Intervention Request, and Voice Check Pass structures.

The assistant will give you dangerous prompts to rewrite, evaluate your rewrites, and help you practice the constraint-first architecture until it becomes instinctive.

Start by typing: "Give me a dangerous prompt to rewrite" — or paste a piece of your writing and describe what you originally wanted to do with it.
Voice-Fidelity Prompt Coach
Lab 3
Welcome to the Voice-Fidelity Prompt Lab. Here we practice transforming dangerous open-ended prompts into voice-preserving structures using the five approaches from Lesson 3.

I can: give you a dangerous prompt to rewrite into a better version, evaluate rewrites you bring me, or work through a real editing task you have — helping you build a proper Preservation Mandate prompt for it.

What would you like to start with?
Module 3 · Lesson 4

The Revision Workflow

A structured process for using AI in revision that keeps your voice intact at every stage.
How do professional writers integrate AI into their revision process without surrendering editorial control?

In 2023, novelist Robin Sloan — author of Mr. Penumbra's 24-Hour Bookstore — published a detailed account of his AI-assisted writing process in a widely circulated newsletter. Sloan described developing what he called a "separation of concerns" approach: he used AI exclusively for what he termed "structural intelligence tasks" — checking for logical gaps in argument, identifying scenes that don't advance the narrative, flagging continuity errors — while keeping all sentence-level prose decisions human. "I never ask the AI to write a sentence," he wrote. "I ask it to tell me when my sentences are doing the wrong job."

Sloan's framework became influential in literary circles because it named a principle many writers had intuited without articulating: AI is more useful as a structural reader than as a sentence writer. The voice lives at the sentence level. The structural problems are where AI's pattern-recognition is genuinely useful and where it poses the least threat to voice.

The Three-Stage Revision Model

Experienced writers who have developed sustainable AI-assisted revision workflows tend to use a three-stage model that separates concerns by level of linguistic granularity. Each stage has different AI risk levels and different voice-preservation requirements.

Stage 1 — Structural Review (AI-appropriate): At this stage you are examining large-scale architecture: argument structure, narrative sequence, section balance, pacing, transitions between major sections. AI is genuinely useful here because structural problems are identifiable by pattern-matching without requiring stylistic generation. The AI isn't writing anything — it's reading for patterns and reporting. Voice risk is low because the AI is not producing prose.

Stage 2 — Paragraph-Level Review (AI-assisted with constraints): At this stage you examine paragraph construction, sentence-to-sentence flow, and local coherence within sections. AI is useful here for identifying awkward transitions, unclear pronoun references, and logical gaps — but only with strict Preservation Mandate prompting. Every intervention request must specify what cannot change. Voice risk is moderate.

Stage 3 — Sentence-Level Review (human-primary): At this stage you are working on individual word choices, rhythm, and the micro-level decisions that constitute your voice. AI should be used sparingly here — and only for specific, bounded tasks like finding homophone errors or checking that a specific technical term is used consistently. AI should never be asked to generate alternative phrasings at the sentence level without explicit constraint that the original rhythm and register must be preserved. Voice risk is highest.

The Voice Drift Audit

After any AI-assisted revision session, professional writers who have developed this workflow recommend a Voice Drift Audit before considering a draft complete. The audit has two components.

The Read-Aloud Test: Read the revised draft aloud, marking any sentence where you hesitate or find yourself rephrasing mid-read. These hesitations are your ear catching voice drift — phrasing that is syntactically correct but rhythmically foreign to your pattern. Investigate every marked sentence for AI intervention.

The Lexical Spot Check: Search the revised draft for any word from your never-list or any word that, while technically correct, is characteristic of AI defaults (common culprits include: "utilize," "leverage," "delve," "nuanced," "multifaceted," "underscore," and "robust"). In 2024, the AI writing detection firm Originality.ai published an analysis of 10,000 AI-assisted documents and identified a list of words that appeared at significantly elevated frequency in AI-assisted versus fully human text. Writers who know this list can scan their drafts for accidental AI vocabulary.

The Originality.ai 2024 Finding

Analysis of 10,000 AI-assisted documents found that the words "delve," "tapestry," "nuanced," "multifaceted," "underscore," "vibrant," "robust," and "leverage" appeared at dramatically elevated frequency in AI-assisted text versus fully human writing. These words are not wrong — they are just statistically characteristic of AI generation defaults. Finding them in your revised draft is a signal to investigate whether AI has intruded at the sentence level.

Maintaining Human Authorship at Scale

The question that becomes acute for writers producing large volumes of work is how to maintain voice consistency when AI is involved in revising ten pieces simultaneously rather than one. The answer from practitioners who have solved this problem is: the voice brief must scale, not just the AI use.

In 2023, the content studio Superpath surveyed 380 professional content writers who used AI in their workflows. They found that writers who maintained a formal voice brief and pasted it into every AI session reported significantly higher satisfaction with voice consistency than those who relied on informal, remembered style preferences. The act of externalizing the voice brief — making it a document rather than a mental state — was the critical variable, independent of how sophisticated the brief was.

The practical implication: as your AI use scales, your voice brief must become more precise, not less. Every new AI capability you adopt is an opportunity for more voice drift, not less — unless your documented constraints keep pace.

Separation of Concerns Robin Sloan's framework for AI-assisted revision: using AI for structural intelligence tasks (identifying logical gaps, pacing problems, continuity errors) while keeping all sentence-level prose decisions human.
Voice Drift Audit A post-revision process combining a read-aloud test (marking rhythmically foreign sentences) and a lexical spot check (searching for AI-characteristic vocabulary) to catch voice erosion before finalizing a draft.
Structural Review Stage The first and safest stage of AI-assisted revision, in which AI examines large-scale architecture — argument structure, narrative sequence, pacing — without generating any new prose.
Core Principle

Voice lives at the sentence level. The further your AI use stays from that level — working on structure, architecture, and logical coherence instead — the less threat it poses to what makes your writing yours. Use AI as a structural reader. Be your own sentence writer.

Lesson 4 Quiz — The Revision Workflow

Five questions · Select the best answer for each.
1. What did Robin Sloan call his approach to dividing AI and human tasks in revision?
Correct. Sloan described his approach as "separation of concerns" — using AI exclusively for structural intelligence tasks (logical gaps, narrative pacing, continuity) while keeping all sentence-level prose decisions human.
Not quite. Robin Sloan called his framework "separation of concerns" — assigning AI to structural intelligence tasks and keeping all sentence-level decisions human. He published this in a 2023 newsletter that became widely influential in literary circles.
2. In the Three-Stage Revision Model, at which stage is AI voice risk HIGHEST?
Correct. Stage 3 — sentence-level review — carries the highest voice risk because this is where the micro-level decisions that constitute voice live: word choice, rhythm, register. AI should be used very sparingly here, only for specific bounded tasks with strict constraints.
Not quite. Voice risk is highest at Stage 3 — sentence-level review — because individual word choices, rhythm, and register are the very substance of voice. Stage 1 (structural review) carries the lowest risk because AI is reading for patterns, not generating prose.
3. What does the Read-Aloud Test in a Voice Drift Audit detect?
Correct. The Read-Aloud Test catches voice drift that the eye misses: phrasing that is syntactically correct but rhythmically foreign to the writer's established pattern. Hesitations and mid-read rephrasing are the ear detecting prose that doesn't belong.
Not quite. The Read-Aloud Test is designed to catch rhythmic voice drift — sentences that are technically correct but feel wrong to the writer's ear because they don't match their established syntactic rhythm. AI-characteristic vocabulary is caught by the Lexical Spot Check, the second component of the audit.
4. According to the Originality.ai 2024 analysis, which of the following is a word that appears at dramatically elevated frequency in AI-assisted versus fully human text?
Correct. "Delve" appeared on the Originality.ai list of words at dramatically elevated frequency in AI-assisted text. Others include "tapestry," "nuanced," "multifaceted," "underscore," "vibrant," "robust," and "leverage."
Not quite. "Delve" is on the Originality.ai 2024 list of words appearing at dramatically elevated frequency in AI-assisted text. The full list from the lesson includes "tapestry," "nuanced," "multifaceted," "underscore," "vibrant," "robust," and "leverage."
5. What did the Superpath 2023 survey of 380 content writers find was the critical variable for voice consistency when using AI?
Correct. The Superpath survey found that externalizing the voice brief — making it a document rather than a mental state — was the critical variable. Writers who pasted their formal voice brief into every AI session reported significantly higher voice consistency satisfaction than those relying on remembered preferences.
Not quite. The Superpath 2023 survey found the critical variable was externalizing the voice brief — having a formal written document rather than relying on remembered style preferences. Writers who pasted their voice brief into every AI session reported significantly higher voice consistency, independent of how sophisticated the brief was.

Lab 4 — The Voice Drift Audit

Practice exercise · Run a full Voice Drift Audit on a piece of AI-assisted writing.

What You'll Do

In this lab you'll practice the Voice Drift Audit from Lesson 4. Paste a piece of text that has been touched by AI (or paste any writing and pretend it has been AI-revised), and the assistant will help you run both components: the Lexical Spot Check for AI-characteristic vocabulary, and a structural assessment of which stage of revision each identified issue belongs to.

You'll also practice writing the Voice Check Pass prompt that you'd use in your own AI workflow to catch drift before it reaches final draft.

Start by typing: "Here is a text I want to audit: [paste text]" — or — "Help me practice a Voice Drift Audit. Give me a sample AI-revised paragraph to work with."
Voice Drift Audit Assistant
Lab 4
Welcome to the Voice Drift Audit Lab. We're practicing the two-component audit from Lesson 4: the Lexical Spot Check (scanning for AI-characteristic vocabulary like "delve," "tapestry," "robust," "leverage," "nuanced") and the Structural Review (identifying which revision stage each problem belongs to and what intervention is appropriate).

Paste a text you'd like to audit, or ask me to give you a sample AI-revised paragraph to work with. Either way, we'll run the full audit process together.

Module 3 — Preserving Voice When Using AI

Module Test · 15 questions · 80% required to pass.
1. Which field of study uses statistical analysis of sentence length, word frequency, and syntactic patterns to identify authorship?
Correct. Stylometry is the quantitative study of literary style using statistical analysis — it's what Patrick Juola used in 2013 to identify J.K. Rowling as the author of The Cuckoo's Calling.
Stylometry is the correct term — the quantitative study of literary style using statistical analysis of word frequency, sentence length, and syntactic patterns to attribute authorship.
2. The gradual erosion of a writer's distinctive style markers through repeated AI-assisted editing is called:
Correct. Voice drift is the term for the gradual erosion of a writer's distinctive style markers through repeated AI-assisted editing, resulting in grammatically correct but stylistically anonymous prose.
Voice drift is the correct term — the gradual erosion of a writer's distinctive style markers through repeated AI-assisted editing, resulting in prose that is technically correct but stylistically anonymous.
3. The University of Copenhagen 2022 research found that even minimal AI polishing reduced which quality in student essays?
Correct. The Copenhagen research found that even one round of AI polishing measurably reduced stylistic variance — essays became simultaneously more grammatically uniform and more stylistically anonymous.
The Copenhagen 2022 research found that even minimal AI polishing reduced stylistic variance — essays became more grammatically uniform and more stylistically anonymous simultaneously.
4. An effective voice brief runs approximately how many words?
Correct. 300–600 words is the effective range — specific enough to constrain AI behavior meaningfully across all six elements, but concise enough for practical use as a system prompt prefix.
An effective voice brief runs 300–600 words — long enough to capture meaningful specificity across all six elements, but short enough to function as a system prompt prefix in AI tools.
5. Which of the following is a well-formed behavioral tonal specification for a voice brief?
Correct. This is a behavioral specification — it describes observable textual actions ("I use 'you' and 'we'") rather than emotional qualities. AI can apply these constraints because they are actionable.
Behavioral specifications describe observable textual actions, not emotional qualities. "I address the reader as a peer, use 'you' and 'we,' and acknowledge uncertainty in first person" gives AI something it can actually apply and check against.
6. What is the "never-list" component of a voice brief?
Correct. The never-list is the explicit prohibition section of a voice brief — words, constructions, and rhetorical moves the writer categorically avoids. It provides hard constraints that override AI defaults and is often more useful than the always-list.
The never-list is the explicit prohibition section of a voice brief — words, constructions, and rhetorical moves the writer categorically avoids, providing hard constraints that override AI defaults.
7. Why is the prompt "Fix the grammar" a voice preservation risk?
Correct. "Fix the grammar" is dangerous because AI treats stylistic choices that deviate from statistical norms — sentence fragments for effect, comma splices for rhythmic flow, unconventional punctuation — as errors to correct, not features to preserve.
"Fix the grammar" is dangerous because AI grammar correction treats stylistic choices that deviate from statistical norms — fragments for emphasis, comma splices for rhythm, unconventional punctuation — as errors to be corrected rather than features to preserve.
8. Stanford HCI 2024 research found that placing style constraints before the task specification (rather than after) produced what improvement in stylistic fidelity?
Correct. The Stanford HCI 2024 study found that constraint-first prompting produced a 23% increase in stylistic similarity to reference text compared to the same constraints placed at the end of the prompt.
Stanford HCI 2024 found a 23% improvement in stylistic fidelity when constraints were placed at the beginning of prompts rather than the end — because AI generates left-to-right and early constraints shape generation from the first token.
9. In the Three-Stage Revision Model, what makes Stage 1 (Structural Review) the lowest voice risk stage?
Correct. Stage 1 carries the lowest voice risk because AI is acting as a structural reader — identifying patterns, flagging logical gaps, noting pacing issues — without generating any new prose that could impose its stylistic defaults on the writer's work.
Stage 1 is lowest risk because AI is reading for patterns and reporting — not generating prose. Voice can only be erased when AI is producing language; a structural reading report doesn't replace any of the writer's sentences.
10. Robin Sloan's "separation of concerns" approach assigns AI to which category of tasks?
Correct. Sloan uses AI exclusively for structural intelligence tasks — checking for logical gaps, identifying scenes that don't advance the narrative, flagging continuity errors — while keeping all sentence-level prose decisions human.
Sloan's "separation of concerns" assigns AI to structural intelligence tasks: checking logical gaps, identifying weak narrative sequencing, flagging continuity errors. "I never ask the AI to write a sentence," he wrote. "I ask it to tell me when my sentences are doing the wrong job."
11. Which of the following words was identified by Originality.ai 2024 as appearing at dramatically elevated frequency in AI-assisted text?
Correct. "Tapestry" is on the Originality.ai 2024 list of words at dramatically elevated frequency in AI-assisted text, alongside "delve," "nuanced," "multifaceted," "underscore," "vibrant," "robust," and "leverage."
"Tapestry" appears on the Originality.ai 2024 list of AI-characteristic words. The full list from the lesson: "delve," "tapestry," "nuanced," "multifaceted," "underscore," "vibrant," "robust," and "leverage."
12. What did the Superpath 2023 survey find was the critical factor predicting voice consistency satisfaction among AI-using writers?
Correct. The Superpath survey of 380 writers found that externalizing the voice brief — making it a document rather than a mental state — was the critical variable, independent of how sophisticated the brief was. Writers who pasted it into every session reported significantly higher voice consistency.
The Superpath 2023 survey of 380 professional content writers found that maintaining a formal written voice brief pasted into every AI session was the critical variable — independent of brief sophistication. Externalizing the voice was more important than how detailed it was.
13. The Animalz 2023 internal analysis compared articles generated with behavioral tonal specifications against articles generated with emotional descriptors. What was the on-voice rating for emotional descriptor articles?
Correct. Articles generated with emotional descriptors alone were rated on-voice by editors at only 34%, compared to 71% for articles generated with specific behavioral tonal specifications.
The Animalz analysis found emotional descriptor articles rated on-voice at 34% — compared to 71% for behavioral specification articles. The gap shows the practical cost of vague tonal language in voice briefs.
14. What is the primary purpose of a Voice Check Pass prompt?
Correct. A Voice Check Pass is a second prompt after initial generation, asking the AI to review what it produced against the voice brief and identify every word, phrase, or sentence that falls outside the stated parameters — giving the writer a precise edit list rather than requiring a full human re-read.
A Voice Check Pass asks AI to audit its own output against the writer's voice brief, identifying every element that falls outside stated parameters. It's a post-generation quality check that maintains editorial control without requiring a full human re-read.
15. Which prompt structure correctly demonstrates the Preservation Mandate approach from Lesson 3?
Correct. This prompt opens with explicit prohibitions before the task, follows constraint-first architecture, and replaces an open-ended improvement request with a specific, bounded intervention — the defining features of a well-formed Preservation Mandate.
A Preservation Mandate places explicit prohibitions first ("Do not alter word choice, sentence structure, or tone"), then specifies a single bounded task. The correct option follows this structure precisely — constraint first, specific task second, no open-ended improvement language.