Course Syllabus Prompt

What this prompt does

A course syllabus prompt turns your course details — topic, level, meeting pattern, texts, and how you plan to grade — into a full draft syllabus: measurable learning outcomes, a week-by-week schedule, a grading table that sums to 100%, and the policies students actually check. You bring the decisions; it does the document.

It's a different job from planning a class. The lesson plan prompt drafts one session — an arc, activities, timing for Tuesday. A syllabus is the semester: the document students receive on day one and consult (or claim to have consulted) for the next fourteen weeks. Carnegie Mellon's Eberly Center frames the syllabus as the place you communicate your course design, built by aligning its components with each other — the outcomes justify the assessments, the assessments explain the grading, the schedule delivers the outcomes. That alignment is exactly what a model is good at holding steady across a long document, and exactly what slips when you write the sections weeks apart.

The prompt below drafts around your decisions. It never makes them for you.

The prompt

Draft a course syllabus from the inputs below. Follow every rule.

**Course title and level:** [E.G. "Introduction to Statistics, second-year undergraduate, no calculus required"]
**Term shape:** [WEEKS, SESSIONS PER WEEK, SESSION LENGTH — "14 weeks, two 75-minute sessions per week"]
**Required texts and materials:** [BOOKS, SOFTWARE, LAB KIT — or "none, readings provided"]
**Assessment plan:** [YOUR COMPONENTS AND WEIGHTS — or "propose one and mark it as a proposal"]
**Institutional policy text:** [PASTE ANY REQUIRED LANGUAGE — accommodations, academic integrity, safety. Write "none provided" if you have none yet.]
**Your course rules:** [ATTENDANCE, LATE WORK, AI USE — the calls you've made]
**Where students struggle:** [THE TOPIC THAT SINKS THEM EVERY YEAR, IF YOU KNOW IT]

Rules:
- Learning outcomes use observable verbs — compute, design, critique, defend. "Understand" and "appreciate" are banned, because no one can grade them.
- Every week names its topic, the outcome it serves, and the reading due. Use week numbers only — no calendar dates, no holidays, unless real term dates were provided.
- The grading table sums to exactly 100%, and every graded component traces to at least one outcome.
- Institutional policy text is reproduced verbatim where it belongs, marked "[institutional text — do not edit]". Never paraphrase it, never draft a substitute for it.
- Policies are written in second person, plain language: what happens when work is late, what counts as collaboration, how AI may and may not be used.
- Anything not provided — office hours, room numbers, exam dates — becomes a bracketed placeholder, never a plausible guess.
- Open with a half-page quick reference: meeting times, contact details, the grading table, and the late-work policy. Students who read nothing else will read this.

How to use it

The verbatim rule on institutional text is the one to respect. Accommodation statements and academic integrity language are usually written by people who chose the words for legal reasons, and a fluent paraphrase can quietly change what they commit you to. Paste the official text and let the prompt place it, not rewrite it.

"Where students struggle" earns its input line. Name the topic that sinks students every year and the schedule gives it two weeks instead of one, or moves it away from midterm crunch — pacing a fresh draft never gets right on its own.

Revision works as well as drafting. Paste last year's syllabus as an extra input, list what changed — new text, one fewer exam, a rewritten AI policy — and the rules still apply, which means the grading table gets re-checked against 100% instead of inheriting last year's arithmetic.

One decision the prompt refuses to make: your AI policy. It will format whatever rule you've chosen, but "decide for me" produces a placeholder, not a policy. That call is yours, and it's the section students will test first.

What the model drafts well, and what stays yours

Example output (partial)

Inputs said: Introduction to Statistics, second-year undergraduate, 14 weeks, two 75-minute sessions a week, one required text, grading split across problem sets and exams, students struggle with hypothesis testing.

Learning outcomes. By the end of this course you will be able to:

1. Compute and interpret descriptive statistics and confidence intervals for real datasets.
2. Design a hypothesis test appropriate to a research question, and defend the choice.
3. Critique statistical claims in published work, including identifying misuse of p-values.

Grading. Problem sets (six) 30% · Midterm one 20% · Midterm two 20% · Final exam 25% · Participation 5%

Week 8 — Hypothesis testing I. Serves outcome 2. Reading: ch. 9. First of a two-week block; nothing new is introduced in week 9 so the concept can settle before midterm two.

Late work. Problem sets lose 10% per day late, to a floor of 50%. Your lowest problem set is dropped. [Office hours: placeholder — not provided]

The two-week block on hypothesis testing came straight from the struggle input — the draft spent the schedule's slack where the difficulty actually is, and said so in the week's own note.

Variations

For a discussion seminar

Add the weekly reading load in pages and a participation rubric input. A seminar syllabus lives or dies on whether "participation: 30%" is defined, so make the prompt spell out what a 30% earns — the rubric generator prompt picks up that section if you need it standalone.

For revising last year's syllabus

Paste the old syllabus and a short list of what changed. Add one rule: output a changelog at the top naming every section that differs from the original. You'll catch the policy you forgot you'd changed before a student catches it for you.

For a school classroom

Teachers drafting a course-expectations document have a second reader: parents. Add the audience line and the policies come out readable at a kitchen table — and drop the outcomes jargon entirely, because "your child will be able to" beats a verb taxonomy at open house.

For an online or asynchronous course

Add your time zone, your response-time promise, and the submission platform. Async courses fail on ambiguity about when — so make the prompt state every deadline with a time and zone, not just a day.

Common pitfalls

• Don't let the model write your accommodation or integrity statement from scratch. A confident, official-sounding draft of institutional policy is still fabricated policy. If you don't have the text yet, ship the placeholder and get it from the source.
• Don't accept "students will understand" in the outcomes, even once. It reads fine and grades as nothing — the ban exists because the model will drift back to it whenever an outcome is genuinely hard to phrase.
• Don't trust any calendar date you didn't provide. Given "14 weeks starting in January," a model will cheerfully invent a spring break. Week numbers are the honest default.
• Try instead: before you ship it, read only the quick-reference block and the late-work policy, the way a student will the night something's due. If those two survive that reading, the rest of the document is in good shape.

Who uses this prompt

• College professors: drafting a syllabus for a new course, or re-aligning an inherited one where the grading table and the outcomes stopped speaking to each other. More in prompts for college professors.
• Teachers: turning course plans into the expectations document that goes home in week one. See prompts for teachers.

Both ChatGPT and Claude handle the length without losing the thread, and the structured input format is why — Anthropic's prompt engineering guidance shows that explicit rules and labeled inputs keep long outputs consistent from the first section to the last. In my experience the section that decides whether the whole document works is the half-page quick reference: students read the grading table and the late policy, skim the rest, and email you about anything those two lines didn't settle.