From Draft to Digital: Using AI in College Writing for Professors

If you teach college writing, you are already grading work that may have started in a chatbot. Guarneri gives you a practical way forward: define your AI policy in plain words, build assignments that ask for process and reflection, and treat detection tools with skepticism. It suits writing instructors, first-year composition coordinators, and faculty development leaders who need language and structures they can use this semester, not someday.

1. 1Early chapters on course design and syllabus language

   Start with your policy, not your panic

   A professor sits down to revise a composition syllabus and hits the same wall every instructor faces now: what do I actually say about ChatGPT? Guarneri's answer is to decide before the semester starts. She lays out the spectrum of stances—from banning AI outright to inviting it into every draft—and shows what each choice demands of you. A ban means you must design assessments a chatbot cannot complete. Full permission means you must teach students to verify and cite what the tool produces. The scene that matters is the professor who never stated a policy, then received a stack of essays with machine-flavored prose and had no ground to stand on. The outcome: the syllabus becomes the first line of defense, written in plain sentences students can quote back.

   The idea

   Ambiguity is what breaks AI-era classrooms, not the technology itself. When students cannot tell whether using a chatbot counts as cheating, they guess, and their guesses vary wildly. A written policy does double duty: it sets expectations and it forces you to decide what writing is for in your course. The surprise is how much pedagogy gets clarified by writing two or three honest sentences about AI.

   Try this: Draft a three-sentence AI statement for your syllabus this week—what students may use, what they must disclose, and what happens if they don't.

2. 2Chapters on academic integrity and AI detection

   Detection tools are not lie detectors

   An instructor runs a student essay through a detection tool, gets a percentage back, and feels the pull to act on it. Guarneri slows the moment down. She explains that these tools estimate patterns, they do not prove authorship, and their errors fall hardest on students writing in a second language. The stakes are a student accused, a record possibly marked, a trust between teacher and learner gone. Her alternative is a conversation: ask the student to walk through their draft history, their notes, their choices. The outcome she aims for is integrity built on evidence and dialogue, not on a number in a sidebar.

   The idea

   A probability is not a verdict. Detection scores can be wrong in both directions, and acting on them alone risks punishing honest students while missing actual misuse. The deeper point is that integrity is a teaching problem, not a surveillance problem. When you ask students to show process—drafts, notes, version history—you get better evidence than any tool offers, and you model the very habits of revision you are trying to teach.

   Try this: Before your next suspicious essay, ask for the student's draft history or writing process notes instead of a detection score.

3. 3Middle chapters on assignment design and the writing process

   Make the process visible again

   The classic take-home essay, Guarneri shows, is now the easiest assignment in the world to outsource. Her redesign starts with the process: annotated bibliographies, proposal memos, in-class freewrites, reflective cover letters where students explain their choices. Picture a student writing a cover letter that says, here is what I changed and why, and the professor reading it and knowing the mind behind the paper. The stakes are real—if you only grade the final product, you are grading something a machine can produce. The outcome: assignments that reward thinking in stages, where the machine can help but cannot replace the student's own decisions on the page.

   The idea

   When the product is cheap, the process becomes the evidence. Asking students to document their choices—what they tried, what they rejected, what they learned—turns every assignment into a record of thinking that a chatbot cannot fake. This surprises instructors who assume redesign means more grading; often it means grading differently, with the reflection doing work the final draft no longer does alone.

   Try this: Add a short reflective cover letter to your next essay prompt: what did you change between drafts, and why?

4. 4Chapters on AI limits, accuracy, and critical reading

   Teach the hallucination

   A student asks a chatbot for sources and receives a list of citations that look perfect—authors, titles, years—except some do not exist. Guarneri turns this failure into curriculum. The classroom scene is students checking each citation, discovering the fabrications, and feeling the ground shift under their trust in the machine. The stakes are the habits of a scholarly mind: if you accept fluent text as true, you are not doing research, you are transcribing. The outcome she wants is a class where verification is reflex, where every quote is traced, every source opened, every claim tested against what the library actually holds.

   The idea

   Fluency is not accuracy. Language models produce plausible text, and plausibility is exactly what makes errors dangerous—they arrive sounding confident and correctly formatted. Teaching students to catch fabricated citations does more than protect them from embarrassment; it teaches the core scholarly move of checking a claim against an independent source. The surprise is that AI's failures make the best lessons in critical reading.

   Try this: Have students fact-check one AI-generated citation list against your library databases and report what they find.

5. 5Later chapters on using AI as a writing tool

   Prompting is a writing skill

   Guarneri reframes the prompt itself as a piece of writing with an audience, a purpose, and revision. The scene: a professor demonstrates in class how a vague request produces vague output, then revises the prompt the way one revises a thesis—adding context, constraints, a role, a target reader. Students watch the output sharpen with each revision. The stakes are agency: a student who can direct the tool is a writer; a student who accepts the first output is a passenger. The outcome is a classroom where prompting is taught, critiqued, and improved like any other draft.

   The idea

   The quality of what a machine gives you depends on the quality of what you ask. That makes prompting a teachable, gradable skill rather than a trick—something with criteria, revision, and craft. What surprises instructors is that teaching prompting strengthens traditional writing instruction: clarity, specificity, and audience awareness matter in both directions. The tool becomes a mirror for how precisely a student can think.

   Try this: In your next class, write a prompt live, then revise it in front of students and compare the two outputs.

6. 6Chapters on authorship, voice, and revision

   Voice is the thing worth protecting

   Guarneri's most personal concern arrives late: what happens to a student's voice when a machine smooths every sentence? The scene is a workshop where a draft that once sounded like a particular person now reads like everyone else—competent, clean, and empty. She asks students to compare their own early drafts with AI-polished versions and name what was lost. The stakes are the entire reason writing is taught: not correctness, but the discovery of what you think and how you sound saying it. The outcome: revision practices that keep the human hand in the text, with AI as a sparring partner rather than a ghostwriter.

   The idea

   Writing instruction has always been about more than clean prose; it is how students find out what they believe. If a tool removes the struggle, it removes the learning. The practical answer is not to forbid assistance but to keep the student's decisions visible—which sentences they kept, which they rewrote, which they rejected. Voice survives when ownership survives.

   Try this: Ask students to submit one paragraph they wrote entirely without AI and mark what they would never let a machine change.

7. 7Chapters on assessment and grading

   Grade the thinking, not just the text

   Guarneri closes the loop on assessment: if AI can produce the artifact, the artifact alone cannot carry the grade. Her scenes are rubrics in revision—points shifting from final polish toward process evidence, oral defenses, in-class writing, and conferences where a student explains their own argument. Picture a professor meeting a student one-on-one, asking them to defend a paragraph's structure, and hearing within a minute whether the thinking is theirs. The stakes are fairness: grades that reflect learning, not access to tools. The outcome is an assessment system that survives the technology because it measures what machines cannot supply.

   The idea

   Assessment is the hinge of the whole book. When you grade only the final text, you invite substitution; when you grade the reasoning behind it, substitution gets hard. Conferences, defenses, and staged submissions are not add-ons—they are the new baseline for what evidence of learning looks like. Instructors who make this shift report less anxiety about cheating because the grade no longer hangs on a document that can be borrowed.

   Try this: Convert one essay grade this term into two parts: the written text and a ten-minute conversation defending it.

College writing instructors, first-year composition coordinators, and faculty development staff who need practical AI-era classroom language and assignment designs. Also useful to any professor whose students submit written work.

FAQ