AI Prompting Tips
Learn how to craft better prompts for more useful, accurate AI responses and more efficient use.
AI Prompting Tips
The better the input, the better the output
Good prompts make all the difference. A vague question often gives a vague answer. A clear, structured prompt helps AI give you useful, accurate results.
Think of prompting like giving instructions to a helpful asistant; be specific about what you need and build your prompt and context patiently. AI works very quickly, but in order to get the best answers, you need to work slowly.
Understanding versus clarity
It is also very important not to confuse understanding with clarity. AI tools can be very helpful whether we understand or do not understand something, but they will struggle a lot more if we are unclear versus being clear.
Whether we are speaking to AI tools about things we understand or not, always be clear by trying to accurately describe what it is that you don't understand. Our prompting tips below can help you to do just that!
Important:
Good prompts save time by reducing confusion, but always review and fact-check what AI gives you. AI tools can make mistakes or create fake references.
Why Prompts Matter
AI tools don’t “know” what you want. They respond to how you ask.
A poor prompt is more likely to produce an incorrect, generic, or limited output.
Whilst a stronger (or clearer) prompt does not guarantee high quality output, it reduces the chances of poor quality output.
Your time is valuable. Clear prompts reduce the need to reproduce work, but you’re still responsible for checking the accuracy of any AI output. Clear and efficient use also helps to reduce environmental impact by reducing the number of prompts required to complete a task.
Five Steps to an Effective Prompt
1. Start Simple: Add Context
AI works better when it knows the background. Include:
- Topic
- Audience (Who is this for?)
- Level (School? University?)
Example:
❌ Weak: “Explain photosynthesis.”
✅ Better: “Can you please explain photosynthesis in simple terms for a first-year undergraduate biology student at a UK university? Please describe the key stages step-by-step and the role each step plays in plant growth and development. Thank you.”
💡 Why it matters:
Context stops AI giving you vague answers. Politeness has been shown to improve the quality of AI output (Yin et al., 2024).
2. Define: Be Clear About the Task
Don’t leave AI guessing, state the purpose of your request.
Summarise? Compare? Outline? Test me?
If you only input “Write about climate change,” you’ll get something too long or off-topic.
Example:
❌ Weak: "Define climate change".
✅ Better: “I am a second-year undergraduate student studying conservation science at a UK university. Please can write a bullet point list of the main causes of climate change? This is to help me conduct my initial research for a report. Thank you!”
💡 Why it matters:
Asking for bullet points and giving the AI model your reasoning ("to help me conduct my initial research") provides a framework that the materials can be produced into.
3. Structure: Set the Format
Tell AI how to present the response. You can define:
- Structure: A table? A graph? Statistical analysis? Paragraphs? Bullet points? Written as a letter?
- Tone: Neutral? Conversational? Upbeat
- Register: Formal? Informal?
Example:
❌ Weak: “Best journals for computer science."
✅ Better: “Can you please recommend five key computer science journals which focus on cybersecurity which are appropriate for an MSc in Computer Science at a UK university? Please present these in a table, listing the name of the journal and their primary editorial focus. Thank you.”
💡 Why it matters:
Stating how you wish to receive the information keeps thoughts organised and
makes it easier for you to assess whether the response meets your needs or not.
4. Make It Engaging: Use Personas (Optional)
Ask AI to “act as” someone. It makes your learning active and engaging.
Example:
❌ Weak: “Act as my tutor and mark my work”.
- Why it is weak: No context, no definition of subject area. No AI tool can provide the effective, growth-focused feedback of your tutors.
❌ Weak: “Present arguments against whether blue is a good colour.”
- Why it is weak: Not clear which side of
the argument it is meant to be arguing for/against, “good” is a vague word.
✅Better:
- Debate partner: “Challenge me with three counterarguments on the argument that “social media improves democracy”.”
- Quiz master: “I am studying for an MSc in Genetics at a UK university. Please test me on 10 key genetics terms relevant for my level of study. Make each question progressively harder. Thank you!”
- Job interviewer: “I have an upcoming interview for a Social Media Marketing Intern with a major UK bank. Please ask me three tricky questions for a marketing interview relevant to this field and give feedback on my responses.”
- Language tutor: “Can you practice basic Spanish with me and correct my answers?”
💡 Why it matters:
Personas turn passive reading into active reading/recall, a key skill for deep learning improving retention.
5. Perfect It: Iterate and Improve
Don’t accept the first answer. Treat AI like a collaborator:
- “Can you make this more concise? Thank you.”
- “Please add two real-world examples.”
- “Thank you. Can you now explain it as a summary in bullet points?”
Tip:
Ask follow-ups like:
- “What’s missing from this explanation?”
- “Can
you explain it in simpler terms?”
Remember:
Just because you like the response, that doesn’t mean the response is correct!
Always fact-check: AI can still produce errors, even after refinements.
Like any collaborator, AI models produce better responses when afforded an average level of politeness (Yin et al., 2024). So, using your pleases and thank yous also has practical benefits!
