L1
·
Quiz
·
Lab
L2
·
Quiz
·
Lab
L3
·
Quiz
·
Lab
L4
·
Quiz
·
Lab
Module Test
Module 4 · Lesson 1

Reading the Machine's Red Pen

How AI grammar and style checkers actually work — and where they silently go wrong.
When a tool flags your sentence as "incorrect," who exactly decided what correct means?

When novelist Cormac McCarthy submitted the manuscript of The Passenger to his editor at Knopf, the text was riddled with sentences that any AI grammar tool would flag: comma splices used as deliberate rhythm, dialogue stripped of quotation marks, subordinate clauses piled seven deep. McCarthy's refusal of conventional punctuation is not error — it is style. Grammarly, run on a passage from the finished book in a 2022 test by The Atlantic, returned 47 alerts on a single page. Every one of them missed the point.

The incident is instructive not because AI tools are worthless, but because understanding what they are trained to optimize for is the prerequisite for using them wisely.

How Neural Grammar Checkers Are Built

Modern AI writing assistants — Grammarly, ProWritingAid, the grammar layer inside Microsoft Editor — are trained primarily on large corpora of edited, published prose: news articles, academic papers, business documents. The model learns statistical patterns: what word sequences appear most often in "clean" text versus text that was later revised. That training signal is excellent for catching genuine errors in functional writing. It is a poor guide to literary style.

The key concept is perplexity. A language model assigns lower perplexity to sequences it has seen often. An unusual syntactic move — an intentional fragment, an inverted sentence, a Latinate periodic construction — registers as high-perplexity and gets flagged. The model cannot distinguish between "I made a mistake" and "I made a deliberate artistic choice that departs from statistical norms."

Core Distinction

AI grammar tools optimize for statistical normalcy, not correctness in the absolute sense. They flag departures from the center of their training distribution. Literary voice often lives at the edges of that distribution — and that is its power.

What These Tools Do Well

Despite the caveats, AI editing tools provide genuine value for specific categories of error. Homophones (their/there/they're), subject-verb disagreement caused by intervening clauses, doubled words ("the the"), and missing closing punctuation are all cases where the model's statistical training aligns well with the writer's intent. A 2021 study by the University of Pittsburgh's Writing Center found that Grammarly caught 72% of grammar errors in undergraduate essays that human peer reviewers missed — a substantial benefit for functional prose.

The strategic writer uses these tools on a second pass, after the stylistic decisions are locked. You run the checker not to learn how to write, but to catch the typos you stopped seeing after your fifteenth read-through. Then you selectively override.

The "Clarity Score" Problem

Many AI tools offer aggregate scores: Grammarly's "Correctness," ProWritingAid's "Readability," Hemingway App's grade-level rating. These scores carry an implicit claim — that writing closer to Grade 6 reading level is inherently better. This is a design choice, not a linguistic truth.

Henry James averages a Hemingway App score of "PostGraduate" with most sentences flagged red. Toni Morrison's Beloved returns a readability grade of 10+ and hundreds of passive-voice flags. The score tells you how far a text deviates from a particular norm of plain style. It tells you nothing about whether that deviation serves the work.

The practical discipline is this: know what each metric actually measures before deciding whether to care about it. "Passive voice" flags catch genuine cases of agency-burying in journalism; they misfire constantly in fiction where the passive is chosen to withhold information or weight a phrase differently.

Perplexity A measure of how "surprising" a text sequence is to a language model. High perplexity often triggers flags — but surprise can be artistic intent, not error.
Training Distribution The body of text a model learned from. Flags reflect departure from that distribution, not universal rules of good writing.
Selective Override The practice of reviewing AI suggestions individually rather than accepting in bulk — the core discipline of AI-assisted editing.
Writer's Discipline

Before running any AI editing tool, write one sentence describing the stylistic choices that define this piece — the moves you will protect. Keep it visible while reviewing suggestions. Every flag is a question: does accepting this suggestion protect or erase a deliberate move?

Quiz — Lesson 1

Reading the Machine's Red Pen · 5 questions
1. What does a language model's "perplexity" score measure in the context of AI grammar tools?
Correct. Perplexity measures how far a sequence deviates from what the model has seen most. High-perplexity text gets flagged — but that can reflect style, not error.
Not quite. Perplexity is a probability-based measure of textual surprise relative to training data, not a direct count of errors or a readability score.
2. The 2022 Atlantic test of Grammarly on a Cormac McCarthy passage is most useful as evidence of which limitation?
Correct. The 47 alerts on one page illustrate that tools trained on statistical norms will flag any sufficiently unusual style, regardless of intent or artistic effect.
Incorrect. The case illustrates the tool's inability to distinguish intentional style from error — not a file-handling or factual-error issue.
3. According to the University of Pittsburgh 2021 study, AI grammar tools were most effective for which type of writing?
Correct. The study found Grammarly caught 72% of grammar errors in undergraduate essays that human peer reviewers missed — exactly the functional prose context where these tools perform best.
Not quite. The Pittsburgh study specifically concerned undergraduate essays — functional, conventional prose where statistical training distributions match the writing goals.
4. What does the Hemingway App's grade-level score actually measure?
Correct. Grade-level scores measure deviation from a specific style norm — plain, short sentences. Henry James and Toni Morrison both score poorly by this measure, which says nothing about their literary quality.
Incorrect. Grade-level scores are purely about syntactic and lexical complexity relative to a plain-style norm. They do not assess ideas, commercial potential, or factual accuracy.
5. What is "selective override" in the context of AI-assisted editing?
Correct. Selective override is the core discipline: treating each AI suggestion as a question rather than an instruction, and deciding case by case whether it serves your goals.
Not quite. Selective override means evaluating and deciding on each suggestion individually — neither bulk-accepting nor abandoning the tool entirely.

Lab 1 — Interrogating the Red Pen

Practice identifying when AI grammar flags serve you — and when they don't.

Your Task

Paste a short passage of your own writing (or one of the examples below) into the chat. Ask the AI to act as an AI grammar tool and flag every "issue" it sees. Then push back: ask it to identify which flags represent genuine errors versus intentional style choices. Explore how it distinguishes — or fails to distinguish — between the two.

Try this: "Here is a passage. Flag every grammar and style issue you can find. Then tell me which of those flags would be wrong to accept if this were literary fiction aiming for a fragmented, urgent voice."
AI Editing Lab
Grammar & Style Analysis
Welcome to Lab 1. Share a passage — your own or any prose you want to examine — and I'll work through it with you. We'll identify what a grammar tool would flag, then interrogate each flag together: genuine error or intentional style move? Let's read the red pen critically.
Module 4 · Lesson 2

Prompting for Structural Editing

Using AI to diagnose pacing, argument flow, and scene structure — before the sentence level.
What if the problem isn't the words — it's the bones?

In early 2023, journalist Casey Newton wrote in his Platformer newsletter about using Claude to read a draft of a long-form technology profile — roughly 6,000 words — and return a structural memo: where does the argument go slack, which anecdotes earn their length, where does the reader lose the thread. Newton reported that Claude's structural notes were "startlingly accurate" on two of three major issues his human editor later identified, and wrong in a third case in a way that was itself illuminating — the AI flagged a digression that turned out to be the emotional heart of the piece.

That case encapsulates the structural editing relationship: AI as a fast first-pass reader whose responses reveal something real about the text, even when the diagnosis is wrong.

What Structural Editing Is

Structural editing — sometimes called developmental or substantive editing — operates above the sentence. It asks: Does this piece have a clear purpose? Does the order of sections serve that purpose? Are transitions doing work? Is the pacing matched to the stakes? In traditional publishing, structural editing is the most expensive editorial service because it requires a reader who can hold the whole piece in mind.

AI systems can approximate this function because large language models process text in large context windows — GPT-4 Turbo and Claude 3.5 Sonnet both handle 100,000+ tokens, roughly 75,000 words. They can "read" a full draft in one pass and return a coherent response about the whole. This is genuinely new. No human beta reader reliably provides structural notes on a 10,000-word draft within minutes.

Prompting Strategies for Structure

The quality of structural feedback depends almost entirely on how you prompt. Vague prompts produce vague feedback. The following prompt types have been documented as effective by writing instructors and authors who have published their workflows:

The Reader Experience Prompt: "Read this draft as an intelligent general reader encountering it for the first time. At the end of each section, describe in one sentence what you now understand to be the piece's central claim, and whether that understanding has shifted since the previous section."

The Tension Map Prompt: "Identify the five moments in this draft where a reader is most likely to either accelerate (lean in) or decelerate (skim). For each, explain what causes that response."

The Missing Thread Prompt: "What question does this piece raise in the first 300 words that it never fully answers? Is that omission intentional?"

Each of these prompts forces the AI to make a specific, falsifiable claim rather than offering vague encouragement. You can then disagree, probe, or redirect.

Falsifiability Rule

A structural editing prompt is well-designed when it could produce an answer you disagree with. If every possible answer would feel useful and non-threatening, the prompt is too vague to teach you anything.

The Limits: What AI Cannot Structurally Edit

AI structural editing has documented failure modes. The most significant: it tends to prefer conventional argument structures. A piece that buries the lede intentionally, or that builds toward a conclusion the reader must construct, or that uses juxtaposition rather than transition — these moves often get flagged as problems when they are features. Editor Sari Botton, writing in Catapult in 2023, noted that when she ran experimental personal essays through AI structural analysis, the AI consistently recommended moving the "main point" earlier. "It was right for most essays," she wrote. "It would have killed the specific essays I value most."

Additionally, AI structural notes treat every draft as if it exists in isolation. They have no memory of your stated intentions from a previous conversation unless you re-supply them. Always open a structural editing session by telling the AI what this piece is trying to do, for whom, and in what form it will be published. Context changes everything about what counts as a structural flaw.

Structural Editing Editing that addresses the organization, argument flow, pacing, and overall architecture of a piece — operating above the sentence level.
Context Window The maximum amount of text an AI model can process in a single pass. Larger windows (100k+ tokens) make whole-draft structural analysis possible.
Falsifiable Prompt A prompt designed so that the AI's answer could be wrong in a specific, checkable way — ensuring feedback has diagnostic value rather than empty validation.
Protocol

Before pasting a draft for structural review, write two sentences: (1) what this piece is trying to do, and (2) what structural move it makes that a conventional reader might misread as a flaw. Paste those sentences as your opening instruction. The AI's subsequent notes will be calibrated to your actual goals.

Quiz — Lesson 2

Prompting for Structural Editing · 5 questions
1. What aspect of a draft does structural editing primarily address?
Correct. Structural editing works above the sentence — on the shape and sequence of the whole piece, not the words within individual sentences.
Incorrect. Those describe copy-editing (spelling/punctuation), line editing (word choice/rhythm), or fact-checking. Structural editing addresses the architecture of the whole.
2. Casey Newton's 2023 use of Claude for structural editing is most notable because it showed that AI structural feedback can be valuable even when:
Correct. Newton noted that the AI's incorrect flag — on a digression that was actually the piece's emotional heart — was itself illuminating. A wrong diagnosis can teach you something real about the text.
Not quite. The key insight from Newton's account was that the AI's incorrect diagnosis of one passage as a problem revealed something valuable about where the piece's stakes really lived.
3. What makes the "Tension Map Prompt" an example of a well-designed structural editing prompt?
Correct. Asking for five specific moments and a causal explanation produces claims you can disagree with — the hallmark of a useful, falsifiable prompt.
Incorrect. The Tension Map Prompt works because it forces specific, falsifiable claims — not general impressions, line corrections, or narrow scope.
4. Editor Sari Botton's 2023 Catapult observation about AI structural analysis revealed which specific bias?
Correct. Botton found AI consistently advised front-loading the main point — advice that would destroy essays whose value lies in gradual, reader-constructed meaning.
Incorrect. Botton's documented observation was specifically about AI's preference for conventional front-loaded argument structure — a real and dangerous bias for experimental essay forms.
5. Why should you always open a structural editing AI session by stating the piece's purpose, audience, and publication context?
Correct. Without stated context, AI treats every draft as if it exists in a vacuum. Context fundamentally changes what counts as a problem — a slow opening may be a flaw in a news piece and an asset in a lyric essay.
Not quite. The reason is simpler and more important: AI has no memory of your intentions, so without stated context every structural judgment is made against a generic default, not your actual goals.

Lab 2 — Structural Diagnosis

Practice using falsifiable prompts to get structural feedback on a real draft.

Your Task

Use this chat to practice structural editing prompts. You can paste a draft (or outline), or work through a hypothetical structure. Start by stating the piece's purpose and any structural move you're protecting. Then use one of the three prompt types from Lesson 2: Reader Experience, Tension Map, or Missing Thread.

Try: "This is a 1,200-word personal essay about my grandmother's kitchen. Its structural move is that the central claim — about immigrant memory — never appears explicitly; the reader must arrive at it. Here is the draft. Apply the Tension Map prompt: identify five moments where a reader accelerates or decelerates, and explain why."
Structural Editing Lab
Structure & Architecture
Welcome to Lab 2. Share your draft or outline and tell me two things first: what this piece is trying to do, and what structural move you're protecting. Then choose a prompt type — Reader Experience, Tension Map, or Missing Thread — and we'll put it to work. I'll give you falsifiable, specific structural notes you can push back on.
Module 4 · Lesson 3

Line-Level Editing: Keeping Your Voice

How to use AI for sentence-level revision without homogenizing your prose into "AI style."
If every writer uses the same tool to polish their sentences, do all sentences start to sound the same?

In 2023, researchers at the University of Pennsylvania published a study in PLOS ONE analyzing 1 million academic abstracts before and after the widespread availability of ChatGPT. They found a statistically significant increase in certain high-frequency words — intricate, commendable, meticulous, pivotal — across abstracts published in 2023 compared to 2021. The words themselves are unremarkable; their sudden ubiquity revealed a shared upstream source: large language models trained on similar corpora, producing similar lexical choices when asked to "improve" academic prose.

The finding does not prove AI editing is harmful. It proves that AI editing, used without critical oversight, produces convergence — a narrowing of the vocabulary and sentence patterns that a field uses to express itself. The same risk applies to literary prose.

The Homogenization Problem

When a language model is asked to "improve" a sentence, it draws on the center of its training distribution. That center is competent, clear, moderately formal, and largely anonymous. It is the prose of a well-edited magazine article — never embarrassing, never distinctive. If you accept its suggestions uncritically across a whole manuscript, you end up with prose that is technically better by standard metrics and considerably less yours.

Literary scholars Naomi Baron and Mark Riedl have both written about this risk: AI writing assistance, at scale, may reduce the variance of published prose over time, making the literary landscape more homogeneous. This is a collective problem, not just an individual one.

Prompting for Voice-Preserving Revision

The solution is not to avoid AI line editing but to prompt it differently. Instead of "improve this sentence," use prompts that foreground your voice as a constraint the AI must work within:

"This sentence is unclear but needs to retain its staccato rhythm and its refusal to explain. Suggest three alternatives that fix the clarity problem without changing those features."

"I use semicolons to create a pause that is heavier than a comma but lighter than a period. Suggest a revision of this paragraph that keeps exactly that weight — do not substitute dashes or conjunctions."

"The word choice here feels weak but the specific quality I'm after is flatness, not vividness. Suggest three alternatives that are flatter and more exact, not more colorful."

Each prompt gives the AI a problem to solve (clarity, precision) while specifying which stylistic features are off-limits for revision. This is the difference between using AI as an authority and using it as a constrained tool.

The PLOS ONE Signal

The appearance of words like meticulous and pivotal in academic writing post-ChatGPT is a measurable signal of what happens when AI suggestions are accepted without interrogation. Watch your own editing sessions for words you would never have chosen but accepted because the AI offered them fluently.

The Three-Pass Line-Edit Protocol

Writers who have developed effective AI-assisted line-editing workflows typically describe some version of a three-pass structure. Pass 1: Read the draft alone and underline every sentence that feels wrong — unclear, weak, or structurally awkward. Identify why in a margin note. Pass 2: For each underlined sentence, write a prompt that names the specific problem and the stylistic constraints the revision must honor. Run the AI on each sentence individually. Pass 3: Read the revised draft aloud. Any sentence that sounds unlike you goes back — either to the original version or to a further iteration with a better-calibrated prompt.

The final pass is the one most writers skip and most need. Reading aloud is the fastest test of whether a revision sounds like your voice or like competent-anonymous prose. The ear is harder to fool than the eye.

Lexical Convergence The documented tendency for AI-assisted writing to cluster around a shared vocabulary, reducing the distinctiveness of individual voices over time.
Constrained Revision Prompt A prompt that names both the problem to be fixed and the stylistic features that must be preserved — preventing AI from defaulting to generic "improvement."
Read-Aloud Test The practice of reading revised prose aloud to detect sentences that sound unlike the author's voice — the fastest quality control for AI-assisted line editing.
Voice Audit

After an AI-assisted editing session, run a vocabulary audit: search your manuscript for words you flagged as AI suggestions and accepted. Ask: would I have reached for this word on my own? If not, replace it with the word you would have chosen — even if the AI's suggestion is technically stronger. Voice is an accumulation of specific choices, not a general quality.

Quiz — Lesson 3

Line-Level Editing: Keeping Your Voice · 5 questions
1. What did the 2023 University of Pennsylvania PLOS ONE study find about AI's influence on academic writing?
Correct. The study found a measurable uptick in specific vocabulary items across 1 million abstracts, consistent with shared AI upstream influence — a documented case of lexical convergence.
Incorrect. The study's key finding was lexical convergence — a measurable clustering of vocabulary around AI-favored words like "meticulous" and "pivotal" in post-ChatGPT abstracts.
2. What is the core risk of accepting AI line-edit suggestions uncritically across a full manuscript?
Correct. AI "improvement" draws from the center of its training distribution — capable but anonymous. Uncritical acceptance produces prose that meets standard metrics while losing what makes a voice distinctive.
Incorrect. The primary risk of bulk acceptance is homogenization — prose becomes competent but anonymous, the voice flattened toward a generic center that serves no individual writer's aims.
3. What distinguishes a "constrained revision prompt" from a simple "improve this" request?
Correct. A constrained revision prompt doubles as both a task ("fix clarity") and a set of stylistic guardrails ("preserve staccato rhythm, refusal to explain"). This prevents default-to-generic improvement.
Not quite. The key is the double structure: name the problem AND specify the constraints. An "improve this" prompt gives the AI maximum freedom to converge on its default center.
4. In the three-pass line-edit protocol, what is the purpose of Pass 1?
Correct. Pass 1 happens before AI involvement — it is the writer's independent diagnosis of where problems exist and why. This prevents the AI from setting the agenda for which sentences need attention.
Incorrect. Pass 1 is the writer's solo diagnostic read — identifying problems and their nature before the AI enters. Running AI tools happens in Pass 2; the read-aloud check is Pass 3.
5. Why is the "read-aloud test" described as the fastest quality control for AI-assisted line editing?
Correct. Rhythm, cadence, and voice are auditory properties first. The ear is harder to fool than the eye when it comes to detecting whether a sentence sounds like you or like a well-polished stranger.
Not quite. The read-aloud test works because voice is fundamentally auditory — the ear detects inauthenticity in rhythm and word choice faster than visual reading of the same text.

Lab 3 — Constrained Revision

Practice writing prompts that fix problems without erasing voice.

Your Task

Choose one or two sentences from your own writing that feel weak or unclear. Describe the problem precisely. Then describe the stylistic feature — rhythm, register, tone, syntax pattern — that must be preserved. Write a constrained revision prompt and submit it. When the AI responds with alternatives, push back: ask it to try again if any suggestion sounds anonymous or generic.

Example: "This sentence is muddy: 'The light, which came in sideways, made everything look temporary, which was something he had always felt anyway.' The problem is the nested which-clauses. The constraint: I want to keep the slow, accumulating rhythm — no short declarative rewrite. Give me three alternatives that clear the syntax without accelerating the pace."
Voice-Preserving Line Edit Lab
Constrained Revision
Welcome to Lab 3. Share a sentence that's bothering you, tell me exactly what's wrong with it, and then tell me the stylistic feature — rhythm, register, syntax pattern, word-level register — that a revision must preserve. I'll offer alternatives designed to solve the problem without erasing the constraint. If any of my suggestions sound generic or anonymous, tell me and I'll try again with tighter guardrails.
Module 4 · Lesson 4

Building Your Editing Workflow

Integrating AI into a complete revision process — from first draft to final polish — without losing authorship.
How do you build a revision system that uses AI as a tool without letting it become the author?

In October 2023, novelist Walter Kirn published an essay in The Atlantic describing his experimental use of AI throughout the revision of a long-form magazine piece. Kirn did not use AI to draft; he used it at three specific revision stages: first to read his structural outline before writing began, second to identify pacing problems in a full first draft, and third to flag clichés in his final polish pass. He retained all authority over actual changes. "The machine," he wrote, "is a mirror I can argue with. That's different from a mirror I simply accept."

Kirn's framework — AI as an interlocutor rather than an authority — describes the disposition that makes AI-assisted editing productive rather than corrosive.

Designing a Layered Revision System

An effective AI-integrated revision workflow operates in distinct layers, each with a different tool set and a different type of question being asked. The layers should not collapse into one another — running a grammar check before structural problems are resolved is a waste; running a structural analysis after you've finalized the line edit may uncover problems you no longer have bandwidth to fix.

Layer 1 — Structural (Draft Complete): Use AI to read the full draft for architecture. Use falsifiable prompts (Tension Map, Missing Thread). Identify problems you will address in the next draft. Do not touch individual sentences yet.

Layer 2 — Line Edit (Structure Locked): Work sentence by sentence on the passages you've identified as weak. Use constrained revision prompts. Run the read-aloud test after each session. Track AI-originated vocabulary and audit it before moving on.

Layer 3 — Copy Edit (Voice Locked): Run grammar and style tools. Accept only error corrections. Override anything that touches rhythm, register, or word choice you have deliberately established in Layer 2.

Layer Discipline

The most common mistake in AI-assisted revision is collapsing the layers — accepting a grammar tool's word-choice suggestion during structural editing, or doing line edits before structure is resolved. Each layer should feel complete before the next begins.

Maintaining Authorship Records

As AI-assisted editing becomes standard, a practical question arises: how do you maintain a clear record of what you wrote versus what AI suggested? Several professional practices have emerged. Version control: save a pre-AI draft with a date stamp so you always have the unmediated version. Session logs: keep a text file of AI prompts you used and what category of change they produced. This is useful both for your own clarity and, increasingly, for disclosure requirements from publishers.

In 2023, major publishers including Penguin Random House and HarperCollins began including AI-use disclosure clauses in author contracts. These clauses typically ask authors to disclose if AI was used to generate substantial portions of the text — not if it was used as an editing tool. The distinction matters, and maintaining records supports your ability to make it clearly.

The Interlocutor Disposition

Walter Kirn's phrase — "a mirror I can argue with" — captures the productive disposition. This means: read every AI suggestion as a question about your work, not an instruction. "This paragraph loses momentum" is not a fact; it is a reading, and readings can be wrong. Your job is to decide whether the AI's reading is accurate about your readers' likely experience, and whether addressing the identified problem serves the piece's actual goals.

The authors who report the most productive AI editing relationships are those who maintain strong prior commitments — clear positions on what a piece is doing and why — before they show it to any AI. Writers who enter revision uncertain of their own intentions tend to accept AI suggestions more readily, because any direction feels like progress. The AI fills an editorial vacuum that should be filled by the writer's own vision first.

Layered Revision A workflow that addresses structural, line-level, and copy-editing concerns in sequence — never collapsing them — with appropriate AI involvement at each distinct stage.
Interlocutor Disposition The stance of treating AI suggestions as readings to evaluate rather than instructions to accept — maintaining the writer as authority over revision decisions.
Version Control The practice of saving pre-AI draft versions with date stamps, preserving the writer's unmediated text and supporting disclosure requirements.
Your Editing Manifesto

Before your next revision session, write three sentences: (1) what this piece is trying to do, (2) which AI tools you will use at each revision layer, and (3) one stylistic feature you will not allow any AI suggestion to change. This "editing manifesto" takes two minutes and preserves the conditions under which productive AI-assisted editing is possible.

Quiz — Lesson 4

Building Your Editing Workflow · 5 questions
1. How did Walter Kirn use AI during his revision process, as documented in his 2023 Atlantic essay?
Correct. Kirn used AI as an interlocutor at three specific stages, making no changes without his own deliberate decision. This is the documented example of layered, author-controlled AI editing.
Incorrect. Kirn explicitly did not use AI to draft. He used it at three specific editing stages while retaining full authority over all actual textual changes.
2. In a layered revision workflow, at which layer is it appropriate to run AI grammar and style tools?
Correct. Grammar and style tools belong in Layer 3 (copy edit), after structural problems are resolved and line-level voice decisions are locked. Running them earlier risks wasting effort on sentences that may be restructured.
Not quite. The three-layer sequence is structural → line edit → copy edit. Running grammar tools before structure and voice are resolved is premature and produces wasted effort.
3. What is the purpose of maintaining "session logs" during AI-assisted editing?
Correct. Session logs help writers track the nature and extent of AI involvement — useful for self-clarity and increasingly required by publishers who have added AI disclosure clauses to contracts.
Incorrect. Session logs serve the writer's self-clarity and support publisher disclosure requirements. They do not give AI memory (which resets per session) or train models.
4. What did major publishers including Penguin Random House and HarperCollins begin adding to author contracts in 2023 regarding AI?
Correct. The clauses target AI text generation — not AI-assisted editing — reflecting publishers' concern about authorship authenticity rather than a blanket ban on AI tools.
Incorrect. The clauses specifically address AI-generated text, not editorial tool use. Publishers drew a line at generation, not assistance — making the distinction between the two legally relevant.
5. What does Kirn's phrase "a mirror I can argue with" mean in the context of AI-assisted editing?
Correct. Kirn's phrase captures the productive stance: AI produces readings, not verdicts. The writer's job is to evaluate whether each reading is accurate and whether addressing it serves the piece.
Not quite. "A mirror I can argue with" describes the interlocutor disposition — AI generates a reading you can challenge, not a rule you must follow or automatically reject.

Lab 4 — Your Complete Editing Workflow

Design and test a layered AI revision system for a real piece of your writing.

Your Task

In this lab, you'll practice the full interlocutor stance: describe a piece you're revising, state your editing manifesto (what the piece does, what layers you'll use AI for, what one feature you'll protect), and then run the AI through one full revision layer — structural, line-level, or copy edit — using the appropriate prompt type. Push back on any suggestion you disagree with. Practice the "argue with the mirror" disposition explicitly.

Try opening with: "I'm revising a 2,000-word personal essay. Its purpose: to argue that grief is not linear without ever stating that claim directly. My protected feature: the present-tense fragments that appear every third paragraph. I want to run Layer 1 structural analysis. Use the Reader Experience prompt: track your understanding of the central claim section by section and tell me where it drifts."
Full Workflow Lab
Layered Revision Practice
Welcome to Lab 4. Begin by stating your editing manifesto: what the piece is trying to do, which revision layer you're working on today, and the one stylistic feature I am not allowed to change. Then share your draft or the relevant portion, and tell me which prompt type you want — Reader Experience, Tension Map, Missing Thread, or a constrained line-edit prompt. I'll give you specific, falsifiable feedback. Argue back whenever I'm wrong. That's the point.

Module Test — AI-Assisted Editing

15 questions · Score 80% or above to pass
1. What is the primary reason AI grammar tools flag intentional stylistic departures in literary prose?
Correct. Grammar tools flag statistical departures from their training distribution — high-perplexity sequences — without any ability to detect whether the departure is intentional.
Incorrect. Grammar tools optimize for statistical normalcy within their training distribution, not for prescriptive rules or author-intent detection.
2. Grammarly was tested on a passage from Cormac McCarthy's The Passenger in 2022. What did that test demonstrate?
Correct. 47 alerts on a single page of a canonical novel demonstrates that tool performance metrics and literary quality are not the same measure.
Incorrect. The test's value is as evidence that grammar tools' statistical norms diverge from deliberate literary style — not that the tools are universally ineffective or that McCarthy made errors.
3. What does "selective override" require of a writer using an AI editing tool?
Correct. Selective override is a discipline of individual evaluation — not bulk acceptance, not blanket rejection, but case-by-case judgment aligned with your specific stylistic intentions.
Incorrect. Selective override means evaluating each suggestion individually — a practice between bulk acceptance and blanket rejection that requires the writer to hold clear stylistic intentions.
4. Why is the large context window of models like GPT-4 Turbo and Claude 3.5 Sonnet particularly relevant to structural editing?
Correct. 100,000+ token context windows mean an AI can process an entire book-length draft in one pass — enabling structural analysis at a scale that was previously impractical for rapid feedback.
Incorrect. Context windows enable full-draft processing in one pass — the prerequisite for meaningful structural analysis. They do not grant internet access or session memory.
5. What makes the "Missing Thread Prompt" an example of a falsifiable structural editing prompt?
Correct. "What question does this piece raise but never answer?" produces a specific claim. The writer can say "you're wrong — it does answer it" or "you're right — and here's why that's intentional." Both responses are productive.
Incorrect. The Missing Thread Prompt is falsifiable because it produces a specific named claim the writer can dispute — not a comparison, a scoped comment, or a numerical score.
6. Sari Botton's 2023 Catapult observation revealed that AI structural analysis has a bias toward recommending which structural move?
Correct. Botton documented AI's consistent recommendation to front-load argument — advice appropriate for news and academic writing but destructive to experimental essays that build toward emergent meaning.
Incorrect. Botton specifically identified the front-loading bias — the AI's tendency to treat delayed revelation as a structural flaw regardless of genre or intentional form.
7. What did the 2023 PLOS ONE study by University of Pennsylvania researchers find about AI's effect on academic writing?
Correct. The measurable uptick in specific AI-favored vocabulary across 1 million abstracts is direct evidence of lexical convergence — a narrowing of vocabulary driven by shared upstream AI influence.
Incorrect. The key finding was vocabulary convergence — specific AI-favored words becoming more statistically common across a massive corpus of abstracts after ChatGPT's release.
8. In the three-pass line-edit protocol, what happens in Pass 2?
Correct. Pass 2 is the AI-engaged pass — writing specific constrained prompts for each sentence identified in Pass 1 and generating revision alternatives. Pass 3 is the read-aloud quality check.
Incorrect. Pass 1 is the solo diagnostic read; Pass 2 is the AI-engaged constrained revision pass; Pass 3 is the read-aloud test. Bulk acceptance is explicitly what the protocol prevents.
9. A constrained revision prompt differs from a simple "improve this" request because it:
Correct. The dual structure — problem + constraint — is what prevents the AI from optimizing toward its generic center. Without the constraint, "improve" defaults to competent-anonymous prose.
Incorrect. The defining feature of a constrained prompt is its dual structure: name the problem AND specify what cannot change. This is what prevents homogenization.
10. The read-aloud test after AI-assisted line editing is most useful for detecting:
Correct. The ear detects inauthenticity in rhythm and word choice faster than visual reading. The read-aloud test is the fastest quality control for homogenization introduced by AI editing.
Incorrect. The read-aloud test's specific function in this context is detecting voice homogenization — sentences that are technically correct but sound anonymous rather than like the specific writer.
11. According to Lesson 4, what is the most common mistake writers make in AI-assisted revision?
Correct. Layer collapse is the identified key mistake: running copy-level tools during structural analysis, or doing line edits before the structure is locked, wastes effort and introduces confusion across concerns.
Incorrect. The specific mistake identified is layer collapse — mixing concerns from different revision stages rather than addressing each layer in sequence with appropriate tools.
12. What was the specific AI use by Walter Kirn at his third revision stage, as documented in his 2023 Atlantic essay?
Correct. Kirn's three stages were: structural outline review, first-draft pacing analysis, and final cliché-flagging — a documented example of purposeful, stage-specific AI deployment.
Incorrect. Kirn used AI for structural outline review (Stage 1), pacing analysis (Stage 2), and cliché-flagging at final polish (Stage 3). Each was a distinct, purposeful use.
13. Publisher AI-disclosure clauses introduced by Penguin Random House and HarperCollins in 2023 specifically target:
Correct. The clauses draw a line at text generation, not editorial tool use — making the distinction between AI-generated content and AI-assisted revision legally and contractually significant.
Incorrect. These clauses specifically address AI generation of substantial text — not editorial assistance. The distinction is legally meaningful and supports writers who use AI as a revision tool.
14. The "interlocutor disposition" toward AI editing means treating AI suggestions as:
Correct. The interlocutor disposition positions AI suggestions as one reader's response — specific enough to engage with, provisional enough to dispute. The writer holds authority over what actually changes.
Incorrect. The interlocutor disposition means treating AI as a discussable partner — its suggestions are readings to evaluate, not instructions to follow or noise to dismiss.
15. Writers who enter revision "uncertain of their own intentions" are most at risk from AI-assisted editing because:
Correct. A writer without strong prior commitments about what a piece is doing will accept AI directions readily — because any direction feels like movement. The AI fills the vacuum of vision that should belong to the writer.
Incorrect. The risk is that uncertainty creates a vacuum which AI suggestions fill — not that the tools behave differently or that documentation becomes problematic. The writer's own vision must come first.