Chatbots
Chatbots use text-based responses to engage in conversation or answer questions. Using these tools typically requires creating an account. Examples include:
- Anthropic’s Claude (strong text and reasoning, handles long docs well, doesn’t use input data for training).
- Google’s Gemini (voice input, image/doc/code understanding, image and video generation, integrates with Google services, might use input data for training future models).
- OpenAI’s ChatGPT (voice mode, image/doc viewing, code execution, excellent image/video generation, Deep Research, good mobile app, might use input data for training future models).
- Microsoft’s Copilot (has many ChatGPT features and is available to Windows users).
- Perplexity (AI-powered answer engine and search assistant).
- Other tools include: Deepseek, Mistral, Grok. Note that the chatbots listed here are most commonly used examples, there are many others available.
Multimodal Generative AI
These tools work across many types of media (e.g., text, images, audio). Examples include:
- Text-to-Image Generators: Adobe Firefly, OpenAI’s DALL-E 2, Midjourney AI, and Stability AI, Stable Diffusion create new or edited images based on prompts.
- Text-to-Audio Generators: OpenAI’s Voice Engine, Meta’s VoiceBox, and Microsoft’s VALL-E create audio clips from text. Some can mimic specific voices or read text in multiple languages.
- Text-to-Video Generators: Adobe Express, Google’s Lumiere, Meta’s Emu Video, and OpenAI’s Sora transform text into video clips or animations.
Specialised AI tools
- For research, coding and writing: NotebookLM, Elicit, GitHub Copilot, Grammarly and Consensus.
- For business intelligence, statistics, math and data visualisation: PowerBI, RapidMiner, Tableau, Wolfram|Alpha and Python.
AI Tool Agents
Generative AI models can be extended with plug-ins or “agents” to perform specific functions, similar to installing apps on a smartphone. Examples include:
- AI-Assisted Search Engines like Microsoft’s Copilot and Google’s Search Generative Experience (SGE), which summarise search results.
- AI Notetakers that record and summarise meeting notes, or even devices that can interpret brain signals into text for those with communication disabilities.
A good place to start exploring AI is by trying ChatGPT, Claude, and Gemini – three of the most widely used chatbots. Each offers different models designed for various tasks, typically in three tiers:
- Fast models for casual chat (e.g. GPT-4o, Claude Sonnet, Gemini Flash)
- Powerful models for advanced tasks (e.g. o3, Claude Opus, Gemini Pro)
- Ultra-powerful models for complex problems (e.g. o3-pro)
Casual models can be used for quick questions or brainstorming, and more powerful ones for tasks like analysis, research, or coding (Mollick, 2025).
The quality of output can depend on what you ask and how you ask it, a process known as prompt engineering. While chatbots are improving, research shows that clear formatting of your requests consistently improves results, though other prompting techniques (such as being polite) have unpredictable effects that vary by specific question (Meincke, Mollick, Mollick, & Shapiro, 2025).
Prompting AI
Click here to open or close the video transcript
Now, here’s where the research on prompt engineering becomes really practical. Clear formatting of your request consistently improves the results. Clarity equals effectiveness. So instead of something a little bit vague, like, help me with my class. Try being really specific, almost to the point of granularity. So think about the task, the format, the voice, and the context that you want that help with. So you might say something like, create three discussion questions for undergraduate biology students about photosynthesis that connects to real world environmental applications. Now, that was a mouthful, but you can see it’s got that context, it’s got the task, it’s got the voice, and it’s got the format.
Now, I learned about specificity with AI tools the hard way. My early attempts at AI prompts were as vague as that example I gave you earlier, help me. And it often led to disappointing results. So when I first started playing with AI, I was disappointed and I almost didn’t continue with it, but I started reading about it. I started learning about effective prompting, and I started understanding that effectively prompting AI was actually making me a better educator, because if I couldn’t explain what I wanted to do to the AI tool, how was I successfully and effectively explaining it to my learners? And so as I’ve been engaging in this process, I’ve come to understand that there are sometimes better ways to explain things to AI, but also most importantly, to my learners. And I feel like I’m growing and becoming better at communication. And I’m an English major. I take pride in already being a good writer and communicator, but I’m growing and learning and becoming better, which is part of the evolution of teaching.
So my encouragement for you is to start simple. Pick one tool, try one focus task. Maybe you wanna develop discussion questions or a specific activity for your learners and build from there. Remember that these tools are improving rapidly, but the fundamental skill, knowing how to communicate clearly about your teaching objectives, hopefully as I’ve just shown, that’s something that you can develop and continue to hone right now.
Think of prompting as having a conversation with a knowledgeable but literal assistant—the more specific and clear you are, the better results you’ll get. Bowen and Watson (2024) suggest the following prompting principles for effective prompting:
- Task: What exactly do you want AI to do? (create, analyze, summarise, explain, draw, write, develop, etc).
- Format: What is the specific output? (blog post, lesson plan, legal brief, jargon-free summary, etc).
- Voice: What style of language if desired? (using academic/friendly/comic/medical language, like a copywriter, historian, Master Yoda, etc.).
- Context: what further context or example can you provide? (suitable for undergraduate students, be serious and friendly at the same time, mention your subject, student level, and learning goals, etc.).
For example:
- Instead of “help with my class,” try “create 3 discussion questions for undergraduate biology students about photosynthesis”.
- Instead of “make quiz questions” try “create 5 multiple-choice questions for first-year psychology students about classical conditioning, with one correct answer and three plausible distractors.”
In the next lesson, you’ll get hands-on experience by trying out chatbots to generate text based on your own prompts.
- Bowen, J. A., & Watson, C. E. (2024). Teaching with AI: A practical guide to a new era of human learning. Johns Hopkins University Press.
- Mollick, E. (2025, June 23). Using AI Right Now: A Quick Guide Which AIs to use, and how to use them. One Useful Thing.
- Meincke, L., Mollick, E. R., Mollick, L., & Shapiro, D. (2025, March 4). Prompting science report 1: Prompt engineering is complicated and contingent.
Discussions
What generative AI tools have you already tried and what was your experience like?
Please share your thoughts and questions in the comments section below.