How to write AI prompts that get better answers
Learn simple ways to ask AI so you get what you need first time, like saying who it's for or showing an example. Then save the requests that work so you can reuse them.
Who it’s for: Anyone who uses AI chat and wants more reliable results.

A prompt is simply the message you type to an AI. Most disappointing answers come from a vague prompt rather than a weak AI. Ask for 'a birthday message' and you'll get something bland. Explain that it's for your sister's fortieth, that she loves gardening and that you want it short and funny, and the answer changes completely.
This course teaches the habits that make that difference. You'll learn to say what the task is, who it's for and how long it should be. You'll add the background AI can't guess, like your goal and your deadline. You'll show an example of what you want, so it copies the style and layout. And you'll ask AI to critique its own draft before you accept it.
Later lessons cover reusable prompt templates for tasks you do every week, and how to ask for tables you can paste straight into a spreadsheet. You'll also learn how to ask the newer 'thinking' models, which work through a problem before they answer. You'll finish by testing two versions of a prompt side by side and keeping the one that works best. The course suits anyone who already uses AI chat and wants more reliable results.
By the end you’ll be able to…
- Add context, audience and format to turn a vague ask into a precise one
- Use one example to make AI copy a tone or layout exactly
- Build a reusable prompt template with blanks for a weekly task
- Get a table or JSON back that pastes straight into a spreadsheet
Short on time? Start with the essentials
These 4 lessons cover the most useful ideas in this course. Each one takes about a minute.
How do I ask AI for what I want?
Rewrite a vague request so it names the task, the audience and the length or format you need.
What should I tell AI first?
Add the background the AI can't guess (your goal, constraints, deadline, who it's for) to get a relevant answer.
How do I get AI to match my style?
Include one or two examples of the output you want so the AI copies the style and structure.
Can AI improve its own answer?
Ask the AI to critique its own draft against your goal, then request a revised version that fixes the weak points.
Every lesson in this course
This course has 12 video lessons of about a minute each, from the basics to expert habits. Each one ends with a question, three takeaways and its sources.
Foundations 2 lessons
How do I ask AI for what I want?Essential
Rewrite a vague request so it names the task, the audience and the length or format you need.
“Ask AI to write about dogs, and you'll get something bland.”
What should I tell AI first?Essential
Add the background the AI can't guess (your goal, constraints, deadline, who it's for) to get a relevant answer.
“AI can't know your boss, budget or deadline unless you tell it.”
Building skills 5 lessons
How do I turn rough notes into an update?
Convert rough fictional notes into a short status update, then check dates, names and commitments before sending.
“Your messy notes can become a clear five-line update in seconds.”
Can AI answer like the right expert?
Assign a role and a target reader so the answer's depth, tone and vocabulary fit the situation.
“Ask the same question as a tax adviser, then as a patient teacher.”
How do I get AI to match my style?Essential
Include one or two examples of the output you want so the AI copies the style and structure.
“One good example often beats ten words describing the style you want.”
How do I keep AI's answers short?
Add clear constraints (length, required points, banned words, reading level) and check the output follows them.
“Ask for “under 80 words, no jargon”, and the answer usually gets better.”
Can AI improve its own answer?Essential
Ask the AI to critique its own draft against your goal, then request a revised version that fixes the weak points.
“You don't have to settle for the first draft.”
Advanced 3 lessons
Can I reuse a prompt every week?
Turn a prompt that worked into a template with labelled blanks, and reuse it for a recurring task.
“Write a good prompt once, and you can reuse it every week.”
How do I ask the 'thinking' models?
Write prompts for reasoning models that state the goal, constraints and success criteria instead of step-by-step instructions.
“With thinking models, you can usually skip “think step by step”. Give the goal.”
Can AI turn messy text into a table?
Ask for a table with named columns, then check a few rows against the original before using it.
“Fifty messy customer emails can become one tidy table. Here's how.”
Expert 2 lessons
How do I ask about a long document?
Organise a long prompt with tagged sections, put documents at the top and the question at the end.
“With a long document, put your question at the end, not the start.”
Which of my prompts works best?
Compare two prompt versions on a fixed set of test inputs, pick the winner and save it to a personal library.
“Your best prompt only counts if it wins on five tests, not one.”
Apps we recommend
These are the tools this course uses, and what each one is best for. We have no partnership with any of them.
ChatGPT
Best for: Practising prompts on everyday tasksFree tier is ideal for practising iterations; projects keep your best prompts close.
FreemiumClaude
Best for: Long, structured prompts with documentsFollows long, structured prompts closely and is strong with documents placed in the prompt.
FreemiumGoogle AI Studio
Best for: Testing structured output and settingsFree playground to compare models, adjust temperature and request JSON output.
FreemiumBuild and test a prompt for a task you repeat
Choose a task you do every week, such as an update email, study notes, a meal plan or a report. Write a template with context, reader, an example and limits, then test it against your old prompt on five real inputs, anonymised where needed, and keep the winner.
What you’ll make: A saved prompt template with labelled blanks, plus a side-by-side comparison on five test inputs.
Check your own work
- The template names the task, the reader, the format and the length
- It includes at least one example of the output you want
- Both versions were run on the same 5 inputs and scored against criteria written beforehand
- Every output was checked against the stated limits, and any misses are noted
- Test inputs contain no real personal or confidential data
Further reading and viewing
We’ve chosen these articles, research papers, videos and guides from other trusted sources.
Prompting best practices
Official techniques: clarity, examples, XML structure, long context and agent prompts.
Prompt engineering
OpenAI's guide to message roles, examples, formatting and prompting reasoning models.
Gemini prompting guide 101
Ready-made prompts for everyday work tasks built on persona, task, context and format.
Interactive prompt engineering tutorial
Nine hands-on chapters with exercises, from basic structure to complex prompts.
Introduction to prompting
Free, beginner-friendly open course covering core and advanced prompting techniques.
Questions about this course
How do I write a good prompt?
Start by stating the task, who it's for and the format you want. Then add any background the AI can't guess, such as your goal, limits and deadline. If you have an example of what good looks like, include it. Finally, read the answer and reply with a follow-up rather than starting again. Small additions like these make a big difference.
What is a prompt template?
A prompt template is a prompt that worked well, saved with blanks for the parts that change. For example, a weekly update template might have blanks for the project, the reader and this week's notes. You fill in the blanks, paste it in and get a consistent result each time. It saves you rewriting the same instructions every week.
Should I give AI an example of what I want?
Yes, when style or layout matters. Describing a tone is hard, but showing one is easy. Paste one or two short examples and ask the AI to match them. It will usually copy the structure, length and voice much more closely than if you only described them. Just make sure the examples don't contain private or confidential details.
Do I need to prompt 'thinking' models differently?
Yes, a little. Reasoning models, often called thinking models, work through a problem before they answer. They tend to do best when you state the goal, the limits and what a good answer looks like, and then let them plan the steps. Long step-by-step instructions can get in their way. For quick everyday questions, a standard model is often faster.
Is the course free?
Your first five lessons in Tau are free. Tau Pro opens every lesson in every course, including the advanced and expert lessons on templates and testing prompts. Each lesson is a short video of about a minute, followed by one question and three takeaways. You can also practise a skill on a realistic task and get feedback from AI.
Get early access to this course
Tau opens soon. Join the waitlist and we’ll email you as soon as you can start this course.
We’ll only email you about Tau, and you can unsubscribe any time. Privacy policy.
We’ll email you as soon as Tau opens.


