Chapter 20 Quiz: Writing and Editing with AI
Test your understanding of AI-assisted writing workflows, voice preservation, and editing techniques. After answering each question, reveal the explanation using the dropdown.
Question 1
Which stage of the writing workflow has the highest AI fit with the lowest risk?
A) First draft generation B) Line editing C) Ideation D) Audience adaptation
Answer
**C) Ideation** Ideation has the highest AI fit with the lowest risk because nothing is committed at this stage. You are exploring angles, not publishing them. A bad AI idea at the ideation stage costs nothing — you simply do not use it. The expand-then-contract approach works particularly well here: generate many more ideas than you need, then select from the best. First draft generation also has high AI fit, but requires more human rewriting to preserve voice. Line editing has a significant risk of AI voice bleed.Question 2
What is "AI voice bleed" and what is the primary technique for detecting it?
A) AI models copying text from their training data; detected by plagiarism checkers B) AI's default stylistic patterns appearing in writing that should sound like you; best detected by reading aloud C) AI making factual errors in writing; detected by fact-checking D) AI changing the length of a document beyond what was requested; detected by word count comparison
Answer
**B) AI's default stylistic patterns appearing in writing that should sound like you; best detected by reading aloud** AI voice bleed refers to the recognizable patterns of AI-generated prose — unusual smoothness and balance, three-part lists, hedging phrases, a characterless polish — appearing in work that should carry the author's distinctive voice. It is best detected by reading the piece aloud, because voice anomalies that are invisible to silent reading become immediately obvious when spoken. The ears are far better than the eyes at catching tone and rhythm problems.Question 3
In the style extraction technique, what are you asking AI to do?
A) Write in the style of a famous author B) Analyze your own writing samples and produce a style guide you can reuse C) Copy your previous blog posts and use them as templates D) Remove AI patterns from a piece you have already written
answer
**B) Analyze your own writing samples and produce a style guide you can reuse** Style extraction is the process of submitting three to five examples of your best writing to an AI, asking it to analyze the distinctive characteristics of your style (sentence length variation, vocabulary register, transition patterns, tone), and then asking it to produce a brief style guide based on that analysis. The resulting style guide can be included in future prompts to steer AI output toward your voice rather than the AI's default style.Question 4
What is the key difference between "AI as ghostwriter" and "AI as editor"?
A) Ghostwriting is legal; editing may raise copyright issues B) As ghostwriter, AI brings its own voice because you have not provided one; as editor, AI works within the constraints of your existing prose C) Ghostwriting produces better output; editing produces safer output D) There is no meaningful difference — both produce the same quality of output
Answer
**B) As ghostwriter, AI brings its own voice because you have not provided one; as editor, AI works within the constraints of your existing prose** The fundamental difference is what provides the voice. When AI is ghostwriting — given a brief and asked to produce prose — it defaults to its own style because there is no human prose in the context window to pull toward. When AI is editing — given your prose and asked to improve it — it works within the constraints of text you have already written. More of your prose in the context window means less AI voice in the output. This is why the "editor" mode is generally safer for voice preservation than the "ghostwriter" mode.Question 5
What does the "expand-then-contract" technique involve?
A) Writing a long first draft, then asking AI to cut it to the target length B) Generating AI output that is longer than needed, mining it for the best elements, then writing a shorter version incorporating those elements C) Starting with a very short outline and expanding it with AI assistance until it reaches the target length D) Asking AI to generate two versions — one expanded and one contracted — and choosing between them
Answer
**B) Generating AI output that is longer than needed, mining it for the best elements, then writing a shorter version incorporating those elements** Expand-then-contract is based on the insight that it is easier to edit than to originate. By generating a draft that is 20-50% longer than your target, you produce more raw material than you need, then select and edit down to the best version. The AI draft is a quarry for ideas, structures, and phrasings — not a finished product. You do not adopt AI's prose wholesale; you use it as a source of options that you then reshape in your own voice. This typically produces better results than asking AI to write exactly to a target length.Question 6
You are asking AI for a developmental edit. Which of the following prompts is most appropriate?
A) "Clean up this draft and make it more polished." B) "Fix all the grammar and spelling errors in this piece." C) "Act as a developmental editor. Evaluate the argument, structure, completeness, and what would make a skeptical reader unconvinced. Do not rewrite anything — only provide editorial feedback." D) "Rewrite this piece in a more professional style."
Answer
**C) "Act as a developmental editor. Evaluate the argument, structure, completeness, and what would make a skeptical reader unconvinced. Do not rewrite anything — only provide editorial feedback."** Developmental editing addresses argument, structure, and completeness — not prose quality. The critical instruction "do not rewrite anything" is what distinguishes a developmental edit request from a line edit or rewrite request. Without this constraint, AI will typically respond to any editing request by improving prose rather than critiquing structure. Options A and D would produce line-level rewrites. Option B is a proofreading request. Only Option C correctly specifies the developmental editing mode.Question 7
At which stage of the writing workflow is AI fit described as "very high" with "low risk"?
A) First draft B) Developmental edit C) Line edit D) Proofreading
Answer
**D) Proofreading** Proofreading — correcting grammar, spelling, punctuation, and capitalization errors — has the highest AI fit and lowest risk of any writing stage. It is mechanical, well-defined, and does not involve creative or voice-dependent decisions. The proofreading request can be made with a simple, constrained prompt that limits AI to error correction only. Line editing has higher risk (voice bleed). First draft generation and developmental editing require more human judgment to manage effectively.Question 8
When using AI to work on long-form content (5,000+ words), what is the most important structural safeguard?
A) Always have AI write the entire document in a single prompt to maintain consistency B) Include the full outline, your introduction, and any previously written sections in each section-generation prompt C) Let AI write all sections first, then edit them all at the end D) Use a different AI model for each section to avoid repetition
Answer
**B) Include the full outline, your introduction, and any previously written sections in each section-generation prompt** Long-form AI collaboration fails most often due to coherence problems — voice drift, internal contradictions, repeated phrases, and loss of argumentative thread across sections generated independently. The fix is providing consistent context: every time you generate a new section, include the full outline, your human-written introduction (which establishes voice and frame), and any sections already completed. This anchors each generation in the same context and dramatically reduces drift. Writing your own introduction is also important — it is the single highest-value human contribution in a long-form AI workflow.Question 9
In the multiple drafts technique, what is the purpose of requesting three different versions?
A) To save time by having AI complete the draft three times as fast B) To force AI to explore the solution space more fully and to give the writer genuine choice among alternatives C) To identify which version is the most accurate D) To test three different AI models against the same task
Answer
**B) To force AI to explore the solution space more fully and to give the writer genuine choice among alternatives** When you ask AI for a single version, it produces one path through the solution space — often a competent but not optimal path. Requesting three genuinely different versions forces broader exploration and gives you real alternatives to choose from. The synthesis step — taking the best elements from multiple versions — typically produces better output than any individual version. This technique also counteracts the psychological trap of treating the first AI output as the "right" output, which it often is not.Question 10
What is the correct rule regarding fact-checking in AI-assisted writing?
A) Fact-check only statistical claims; narrative descriptions do not require verification B) Fact-check AI-generated content but not your own contributions, since those are your professional knowledge C) Every factual claim in AI-assisted writing must be verified against a primary source before publication D) Fact-check only in high-stakes documents; internal communications and blog posts carry lower risk
Answer
**C) Every factual claim in AI-assisted writing must be verified against a primary source before publication** The fact-checking rule is absolute, not conditional. AI produces false factual claims with the same confident tone as accurate ones, and the errors can be difficult to detect because they are plausible-sounding. The verification must happen before publication — not "when you have time" and not selectively for certain types of claims. The embarrassment cost of a published fabricated statistic or misattributed quote typically exceeds the time saved by skipping verification. Treat AI's factual claims as unverified until you have checked them.Question 11
Alex's blog post workflow in Section 10 takes approximately 90 minutes total. What does she do herself that AI does not do?
A) Everything — she writes the whole post and uses AI only for proofreading B) Nothing — the workflow is entirely AI-generated C) The introduction, key revisions throughout, two specific clunky sentences for alternatives, and the final read-aloud pass D) Only the final proofreading pass
Answer
**C) The introduction, key revisions throughout, two specific clunky sentences for alternatives, and the final read-aloud pass** Alex uses AI for ideation, outlining, first draft generation, developmental editing feedback, and proofreading. However, she writes her own introduction from scratch (the AI version lacks the specific hook she has in mind), she rewrites sections where AI made unsupported generalizations, she does most of her own line editing, and she reads the final piece aloud as her non-negotiable final step. The read-aloud step is particularly notable — it is her primary voice-check method and catches problems that silent reading misses.Question 12
What is "AI padding" and how do you address it?
A) AI adding unnecessary visual spacing in documents; fixed by reformatting B) AI filling space through redundancy — repeating points in slightly different language, adding qualifiers that add no information; fixed by aggressive cutting C) AI producing output that is shorter than requested; fixed by asking for more detail D) AI inserting placeholder text where it lacks information; fixed by providing more context
Answer
**B) AI filling space through redundancy — repeating points in slightly different language, adding qualifiers that add no information; fixed by aggressive cutting** AI padding is one of the most common AI writing failure modes. When given a target length, AI often reaches it through redundancy rather than genuine content: restating points in different words, adding transitional summaries that recap what was just said, inserting qualifier clauses that hedge without adding information. The fix is editing for concision aggressively — treating the AI draft as raw material that probably needs 20-30% cut, not a polished piece that needs minor refinement.Question 13
What does the research evidence suggest about AI writing assistance and writing quality for expert writers?
A) Expert writers see the largest quality gains from AI assistance B) AI assistance consistently improves writing quality across all skill levels equally C) For expert writers, AI assistance can actually introduce quality regressions if the workflow is not carefully managed D) Expert writers should not use AI assistance because it always degrades quality
Answer
**C) For expert writers, AI assistance can actually introduce quality regressions if the workflow is not carefully managed** The research evidence suggests that AI writing assistance improves quality most significantly for lower-skilled writers, who benefit from AI's structural competence and breadth. For expert writers — who already have strong structural judgment and a distinctive voice — poorly managed AI assistance can degrade quality by introducing AI voice bleed, flattening idiosyncratic but effective stylistic choices, and producing the kind of smooth competence that lacks the edge of genuinely expert writing. This is why workflow design matters more than model capability, and why voice preservation techniques are especially important for professional writers.Question 14
Elena's consulting report scenario illustrates a specific choice she makes at the executive summary stage. What is it, and why?
A) She asks AI to write the executive summary first, before any other sections B) She writes the executive summary herself, because she has been burned before by AI summaries that sounded generic and lost the sharp analytical point of the research C) She uses a different AI model for the executive summary than for the body sections D) She does not include an executive summary because clients do not read them
Answer
**B) She writes the executive summary herself, because she has been burned before by AI summaries that sounded generic and lost the sharp analytical point of the research** Elena's decision to write the executive summary herself is based on experience, not policy. She has found that AI executive summaries tend toward generic synthesis that loses the specific analytical insight that differentiates high-quality consulting work. The executive summary is often the only thing a client reads — and if it fails to convey the core analytical point, the entire engagement's value is diminished. This is an example of a sophisticated workflow choice: not blanket avoidance of AI, but deliberate allocation of human effort to the highest-stakes, most voice-dependent outputs.Question 15
When should you NOT use AI assistance for email writing?
A) When the email is to an external stakeholder B) When the email requires an apology C) When the email is brief and casual, and AI overhead would exceed the writing time D) When the email conveys negative information