Prompt engineering is a complex, delicate art form that requires sophisticated training and ongoing, dedicated effort. Even then, only a prestigious few can hope to master advanced AI prompting techniques… Sike.
Writing effective prompts is actually pretty simple if you’re clear, specific, and provide the chatbot with context.We cover the basics below and reveal the latest prompting technique, Quad-shotting, and you can skip to the prompt examples, templates, and tips.
What is a prompt?
A prompt is the message you type into a chatbot or other generative AI to tell it what you want. Prompts can be a question, requesting help, or giving it a task like summarizing text, generating code, and even translating text to other languages.
What is prompt engineering?
Prompt engineering is just how to write your requests to get the response you’re looking for from a chatbot. “Engineering” makes it sound like a really technical (sometimes it can be), but for most educators’ use cases, you just need to be clear and specific about what you want and provide details to help the chatbot tailor its response.
What is the best AI chatbot?
The best AI chatbot really depends on what you’re doing. If it’s a general writing task, ChatGPT and Claude are likely your best bet. If it’s technical writing, Gemini and DeepSeek Chat are solid options. You can compare AI chatbots here.
Types of prompting techniques + unnecessary terms
There are a ton of terms that make prompting seem like an extremely complex and technical process, but they mostly just describe things people naturally do when they use AI.
Take shot prompting, for example. It includes Zero-shot, One-shot, and Few-shot. It just refers to the number of examples you provide in your prompt that show the chatbot what you want.
- Zero-shot = 0 examples
- One-shot = 1 example
- Few-shot = a few examples
Then there’s prompt chaining and chain of thought prompting, which sound similar but are different, and neither is overly technical.
- Prompt chaining means giving the chatbot one prompt at a time instead of everything at once. The idea is that each answer helps with the next step/response.
- Chain of Thought (CoT) prompting asks a chatbot to show its reasoning before answering. It can be as simple as adding, “Explain your steps before answering,” to the end of a prompt.
Not so daunting, right? Not so fast. Here’s the real expert revealer…
Have you heard of Quadrillion-shot prompting, which is the latest and greatest prompting technique? Quadrillion-shot prompting, also called Quad-shotting, is when you use a quadrillion examples in your prompt. We just made it up to show how unnecessary some prompting terms are, especially the ones in the dropdown below.
Thread of Thought (ToT) prompts just include something like, “Walk me through how [topic/subject] works in a step-by-step manner while summarizing each one.”
Contrastive Chain of Thought (CCoT) prompting gives the chatbot correct and incorrect examples/reasoning.
Generated Knowledge prompting is when you ask the chatbot about a topic and then ask it to answer a question based on its response.
- Prompt 1: Explain how [topic/subject] works and create a list of key concepts.
- Prompt 2: Using the information you provided, explain the steps to [related aspect of topic/subject].
Meta-prompting is basically just asking the chatbot to create its own prompt (or improve the one you provided), then following up with another prompt asking it to answer the prompt it created or improved.
Chatbot prompting tips for educators:
Provide specific context and clear instructions
Providing specific, relevant details can improve the chances of getting the response you want.
Think of writing a prompt like writing a grocery list for someone who’s shopping for you. The chatbot will usually get the items on your list, but if you want something specific—like a brand, flavor, or amount—you have to say so.
If you want vanilla ice cream, adding “snacks” to your list doesn’t really help. Even adding “ice cream” might get you the wrong flavor. Be specific: ½ gallon of Haagen-Dazs Vanilla Ice Cream.
Details to consider providing when writing prompts for AI chatbots:
- Voice: Who should it write like? e.g., “Write as a historian” or “Write like [specific person]
- Tone: Casual, formal, humorous, academic, dramatic?
- Audience: Who’s reading this? Students? What kind—third graders, college freshmen, PhD researchers? What’s their major? How much do they already know?
- Format: How should the response be provided to you? Paragraphs or bullet points? If you want a chart or table, be specific about the type, how many rows and columns, titles, etc.
- Length: How many words, sentences, paragraphs, bullet points, etc.?
- Context: Any key details? What information and nuances are needed?
- Examples: Provide good and bad examples so the chatbot understands what you want.
AI prompt examples and templates
Use these prompting examples and templates as a starting point, and tailor your prompt with information specific to your assignments and assessments.
Don’t overthink it
- Don’t worry about order: Chatbots use machine learning to process your whole prompt at once, so the order of the information you provide usually doesn’t matter that much in most cases. Just focus on providing the right information.
- Write naturally: Chatbots can “understand” everyday language (even our slang and sarcasm) thanks to a technology called natural language processing, which is part of how large language models (LLMs) work. So, you don’t need to use special wording or commands. Just write to it like you would a real person, and it’ll usually get the gist of what you’re saying.
Remember chatbots’ original purpose and fact-check everything
Chatbots are used for many things, but their original purpose is still the same: to chat with you.
You might ask them to explain complex topics, check facts, or even solve math problems—but at their core, chatbots are built to converse. They aren’t inherently designed to conduct research, verify information, or solve math problems. They can help with those tasks, and some even have features like “Deep Research,” but those are just extensions of their original purpose.
That said, always, always, always fact-check chatbot responses because they don’t generate text based on facts. They just predict what word is most likely to come next.
Treat all conversations with AI like public conversations
Regardless if it’s a public or private chatbot, don’t include any personally identifiable information (PII) or sensitive organizational details in your prompts or files you upload (Excel, Word, etc.).
Anonymize information to protect learners, yourself, and your organization. For example, use “University/Company ABC” instead of your company name, or “Student/Employee A” instead of specific names.
Control AI use during exams and assignments
Generative AI chatbots have plenty of benefits for educators and learners.
They can serve as a smart assistant that helps educators develop course materials, personalize learning experiences, and even improve accessibility. When learners use them appropriately—which means not overly relying on them or simply copying and pasting the responses as their own—chatbots can increase engagement, improve learning and retention of information, and sharpen critical thinking and problem-solving skills. That said, when they’re overused, the exact opposite is true.
The good news? You can control if, when, and how learners use AI during online exams and assignments by using hybrid virtual proctoring. It allows you to block AI entirely or allow it at certain times. So, whether it’s a traditional online exam, a written assignment, or even a math test (online or handwritten), you’ll know learners didn’t use AI.
More Reading & Learning
