AI-ready syllabus: a section-by-section guide

The AI policy paragraph is the easy part. Syllabus briefings gives you three ready-to-use versions at different permissiveness levels. This guide covers the rest of the syllabus, because AI touches more sections than the policy slot, and because a common piece of inherited template language actively contradicts any AI policy you adopt.

Everything here is generic. Adapt it to your institution’s template, and keep every statement your institution requires verbatim. Nothing in this guide overrides institutional policy.

Start with the collision check

Most syllabus templates carry language written before generative AI, and some of it contradicts an AI policy sitting two sections away. Before adding anything, check four places:

  1. The academic conduct section. Language like “all assignments are to be completed individually and without any study aid, including online sites” is a blanket that forbids what your AI policy permits. A student reading both is asked to resolve the contradiction alone.
  2. The required materials section. If the policy permits or expects an AI tool, the materials section should say which one, at what cost, and with what free alternative.
  3. The grading section. If some assignments permit AI and some do not, the grade formula should show which components are which.
  4. Honor pledge or proctoring language. “No unauthorized aid” only works if the syllabus says what is authorized, assignment by assignment.

The repair pattern for all four is the same: condition the blanket on the assignment’s stated category rather than deleting it.

1. The AI policy statement

Adopt or adapt one of the three paragraphs in briefings.md as the course default, then add the sentence that makes it operational:

Each assignment states which of four categories applies: no AI, AI as tutor, AI as collaborator, or AI as object of study. The assignment’s stated category takes precedence over this course default.

Labeling per assignment is what makes the policy enforceable and fair: students are never guessing, and two assignments with different purposes can carry different rules. The assignment templates carry the label structure.

2. Academic integrity and conduct

Keep your institution’s required language. Where the template offers suggested language with a pre-AI blanket, condition it:

Unless an assignment’s AI category states otherwise, all work is to be completed individually and without outside aid. Where AI use is permitted, you are responsible for verifying the output, disclosing your use as the assignment requires, and submitting only work you can explain.

State how the course reads integrity, so the standard is visible before anyone is graded:

This course assesses your work through the evidence of your thinking: your process, your explanations, and your ability to answer questions about what you submitted. It does not rely on AI-detection software.

3. Your own AI use

The section most templates lack entirely. A syllabus that sets AI rules for students while saying nothing about the instructor’s own use is structurally fragile, and students notice. One line covers standing practice:

I use AI in preparing parts of this course, including [drafting materials / generating practice problems / first-pass feedback, adjust to fit]. Where AI materially shaped something you receive, I will say so on that item. The disclosure standard I ask of you is the one I hold myself to.

4. Course goals and objectives: add the scope note

Alongside what the course covers, state what the course is for and where it sits inside its discipline. An AI answers from the middle of a discipline, not from the position your course occupies in it, so a correct answer by the wrong method is a live failure mode that no topic list catches. A skeleton:

This course approaches [subject] as preparation for [what comes next], and it emphasizes [orientation: the methods that scale, proof technique, model criticism, adjust to fit] over [what it deliberately postpones]. When you use an AI tutor, give it this paragraph.

The last sentence is not decoration. The scope note does double duty: the Study Partner Protocol carries a matching scope field, so students paste your paragraph straight into their tutor setup and every student’s AI holds the same boundary.

5. Grading: split one component, move nothing else

You do not need a new grading scheme. Take one component of the formula you already use and split it. If homework is 20 percent, make it 10 percent AI-permitted and 10 percent independent, and leave every other number alone. The split makes the categories visible in the grade, gives students practice in both modes, and costs no restructuring.

Two supporting lines worth adding:

Each assignment’s AI category appears in the course schedule and on the assignment itself.

Where an assignment asks you to show evidence of your thinking, a weak first attempt at presenting that evidence is treated as missing instruction, not missing ability. I will show you a worked example before any are graded.

The assignment templates and rubrics grade the judgment rather than the product; the assessment models give low-stakes ways to see unaided competence alongside the AI-permitted work.

6. Required materials: the tool, the tier, and the boundary

If the course expects AI use, treat the tool like any other required material: name it, name the tier, and give a no-cost path, since paid tiers are not equally available to every student. Then state the upload boundary:

When working with an AI tool, give it the course scope note rather than uploading course materials. Do not upload the textbook or other copyrighted material, other people’s work without permission, or anything private. If you want an AI aligned to a full text, use an openly licensed one.

7. Attendance and office hours: say what they are for now

Both sections predate AI and both have a new job worth stating. Class meetings are where the course sees your unaided thinking early, while there is still time to act on it, which is the strongest attendance carrot available. Office hours are the human rung of the help ladder:

When you are stuck, the order that serves you best is: your own attempt, then the AI tutor within the course scope, then office hours. Bring the AI conversation with you; where you got stuck with the tool is exactly what I want to see.

8. The schedule: two rows that do the setup

Two schedule entries make the whole policy operational:

Support services

One sentence connects AI use to the tutoring and writing support your institution already offers:

If you work with a tutor or the writing center, bring your AI disclosure (template) so they see the same record of your process that I do.

Adapt freely. These materials are MIT-licensed; the framework behind them is Part II of Learning with AI (SIAM). Corrections and suggestions are welcome through FEEDBACK.md.