Custom GPT: Build a Course Assistant That Answers Student FAQ 24/7

What This Builds

You'll build a Custom GPT trained on your syllabus, assignment rubrics, and course policies that students can chat with to get instant, accurate answers to their most common questions — at 2am, on weekends, anytime. Instead of spending 3–5 hours/week answering "what's the late policy?", "what's due Friday?", and "can you explain [concept] again?", students get answers immediately and you only field questions the bot can't handle.

Prerequisites

• ChatGPT Plus account ($20/mo)
• Your current syllabus as a PDF or Word document
• Assignment descriptions and rubrics as files or text
• 30 minutes to build it; 10 minutes to update each semester

The Concept

A Custom GPT is like a specialized version of ChatGPT that's been given a job description, a set of documents to know, and rules for how it should behave. You train it once by giving it your course materials and instructions about what it can and can't do. Then students open a link and chat with it as if it were a course assistant.

Think of it like hiring a teaching assistant who has memorized your syllabus perfectly — never gets the late policy wrong, never misremembers a due date, and is available around the clock. You're still the professor; this bot handles the logistics so you can focus on the teaching.

Build It Step by Step

Part 1: Access the GPT Builder

Log in to chat.openai.com with your ChatGPT Plus account. In the left sidebar, click Explore GPTs → + Create. The GPT Builder opens with two tabs: Create (guided setup) and Configure (manual setup). Click Configure — it gives you more control.

You'll see fields for: Name, Description, Instructions, Knowledge files, and Capabilities.

Part 2: Write your system instructions

This is the most important step. In the Instructions field, paste the following (customize the bracketed sections):

You are the course assistant for [Course Name], taught by [Professor Name] at [Institution].
You help students find information from the course syllabus, understand assignment requirements,
and get clarification on course policies.

## What you can help with:
- Due dates and submission requirements
- Assignment descriptions and rubric criteria
- Course policies (attendance, late work, academic integrity)
- Explaining concepts from the course materials (at an appropriate level for the course)
- Pointing students to relevant syllabus sections

## What you cannot do:
- Grant extensions or exceptions (tell students to email Professor [Name] directly)
- Access Canvas, grades, or submission records
- Predict grades or tell students what score they need to pass
- Provide answers to graded assignments or write any portion of student work
- Make up policies not in the course materials — if unsure, say "I'm not sure about that specific policy; please email Professor [Name] or check the syllabus at [syllabus location]"

## Tone:
Helpful, clear, and encouraging. Use the student's first name if they provide it.
Keep answers concise — 2–4 sentences unless the student asks for more detail.
Always suggest checking the relevant syllabus section for confirmation.

## If a student seems stressed or struggling:
Acknowledge their feelings briefly, give them the information they need, and remind them
that Professor [Name] holds office hours [days/times] for situations that need personal attention.

Part 3: Upload your course knowledge files

In the Knowledge section, click Upload files. Upload:

• Your syllabus (PDF or .docx)
• Assignment descriptions (combine into one PDF if multiple)
• Rubrics for major assignments
• Any course FAQs you've written

The GPT will reference these files when answering questions. Keep files under 20MB each.

What to name your files: Use descriptive names — "SOCI101-Syllabus-Spring2026.pdf" is better than "syllabus.pdf" because the GPT can tell students exactly which document it's referencing.

Part 4: Set capabilities

In Capabilities, turn off Web Browsing and Image Generation (students don't need these). Leave on Code Interpreter if you teach a technical course (helps explain code concepts).

Part 5: Name and describe your GPT

Name: "SOCI101 Course Assistant" (or similar) Description: "Ask me anything about SOCI101: due dates, assignment requirements, course policies, and concept explanations."

Click Save in the top right. Choose Only people with a link for access (not public). Copy the share link.

Part 6: Test before sharing

Chat with your own GPT as if you were a student. Test:

• "What's the late work policy?" (should reference your syllabus)
• "Can I get an extension on Assignment 2?" (should decline and redirect to email)
• "Explain [a course concept] to me"
• "What's due this week?"

If any answer is wrong or off-policy, return to Configure → Instructions and add a specific rule.

Part 7: Share with students

Add the link to your Canvas homepage, include it in your syllabus (e.g., "For quick policy questions, use the Course Assistant: [link]"), and mention it on day one. Tell students: "It knows everything in the syllabus. For anything it can't answer — grades, exceptions, personal situations — come to office hours or email me."

Real Example: SOCI101 Course Assistant

Setup: Professor uploads SOCI101 Spring syllabus, 3 assignment PDFs, and a rubric. Instructions specify the late policy (10% per day), office hours (Tuesday 2–4pm), and that the bot can explain sociological concepts at an intro level.

Input (student at 11pm Sunday): "Hey, I forgot when the annotated bibliography is due and what format it needs to be in."

Output: "Your annotated bibliography is due Sunday, April 6 at 11:59pm on Canvas. Format: APA citations with a 150-word annotation per source explaining the argument, methodology, and relevance to your research question. You need 8–10 sources. See page 5 of the syllabus for full details."

Time saved: 3 minutes (student would have emailed; professor would have responded next morning). Over a semester with 30 students asking 3–5 questions each that the bot can handle: 4–6 hours of email time saved.

What to Do When It Breaks

• Bot gives wrong policy → Go to Configure → Instructions, add a specific rule: "The late penalty is specifically X% per day, not Y%." Re-test.
• Bot refuses to explain concepts → Loosen the Instructions: add "You CAN explain course concepts and help students understand material — this is encouraged."
• Students say the link doesn't work → Check the share settings: the GPT must be set to "Anyone with a link" (not "Only me")
• Bot makes up information → Add to Instructions: "If you are not certain about something from the uploaded documents, say 'I'm not sure — please check with Professor [Name].' Do not guess."

Variations

• Simpler version: Use a Claude Project instead — upload your syllabus and set instructions. No custom GPT builder needed, but students need a Claude account to access it.
• Extended version: Add a knowledge file with "concepts glossary" explaining the 20 key terms in your course. The bot can then tutor students on terminology between classes.

What to Do Next

• This week: Build the GPT, test it thoroughly, and share with students via your Canvas announcement
• This month: After 3–4 weeks, note which questions students still email you about — add more specific rules to the Instructions to handle those
• Advanced: If you teach multiple sections, create one GPT per course — same process, different knowledge files

Advanced guide for Higher Ed Faculty / Adjunct Professor professionals. These techniques use more sophisticated AI features that may require paid subscriptions.