If you’ve used Claude, ChatGPT, or another AI assistant, a “prompt” is simply the request you give it — a question, instruction, or task. And the quality of that request has a major impact on the quality of the answer you get back.
Strong prompts are usually built from three ingredients: instructions, context, and examples. If you use AI regularly, you may already include some of these without thinking about them. Learning what each one does makes it much easier to improve a prompt when the result isn’t quite right.
This guide breaks down the role of each ingredient, what happens when one is missing, and how to spot the problem quickly instead of rewriting your entire prompt by guesswork.
If you’re new to prompting, start with our complete prompting guide first.
Why This Distinction Matters
A vague prompt or disappointing response can have several different causes.
Sometimes the task wasn’t clear. Sometimes the task was clear, but the AI didn’t have enough background information. And sometimes both were in place, but the desired style or structure was never shown.
These are different problems, and they need different fixes.
Treating every weak response as a sign that “the prompt needs more detail” often leads to longer prompts filled with information that doesn’t solve the real problem.
A better approach is to diagnose the gap first. Look at what went wrong in the response, then decide whether the prompt needs stronger instructions, more context, or an example.
Instructions: What They Do
Instructions define the task. They tell the AI what you want it to produce and can also set the format, length, scope, or other requirements.
The key question is:
What should the AI do?
What happens when instructions are unclear:
The response may address the general topic without completing the task you had in mind.
For example, ask:
Help with our pricing page.
You’re likely to get broad suggestions about pricing pages.
Instead, ask:
Rewrite this pricing page section to lead with value instead of features, in under 100 words.
Now the AI has a specific job to complete.
How to fix it:
Frame the prompt around an action rather than a subject. Add any important requirements around format, length, or scope.
“Help with X” gives the AI a topic.
“Do Y to X, in this format, at this length” gives it a task.
Context: What It Does
Context is the background information the AI needs to tailor its answer to your situation.
Without it, the response may be technically correct but too generic to be useful.
The key question is:
What does the AI need to know about my situation to give me a relevant answer?
What happens when context is missing:
Ask:
How should I price this product?
You’ll probably receive general pricing advice.
Now add:
We’re a two-person business selling to enterprise clients with a six-month sales cycle.
That information changes the kind of answer you need.
How to fix it:
Include the facts that could meaningfully change the response, such as:
- your business type
- your target audience
- important constraints
- your goals
- what you’ve already tried
A useful rule of thumb is this: if removing a detail wouldn’t change what a good answer looks like, you probably don’t need it. If removing it would change the recommendation, it belongs in the prompt.
Examples: What They Do
Examples show the AI what you want rather than relying entirely on description.
They’re particularly useful for things that are open to interpretation, such as tone, structure, level of detail, or writing style.
The key question is:
What should the finished result look or sound like?
What happens when an example is missing:
Your instructions and context may both be strong, yet the final result still feels wrong.
For example, phrases such as “professional but approachable” can mean very different things to different people. A short sample written in the tone you want gives the AI a much clearer target.
How to fix it:
Include one representative example, such as:
- a sentence written in the right tone
- a past email or piece of copy you liked
- a structure you want the output to follow
- a competitor example for format or approach, while using your own original content
One good example is often enough. Adding too many can make the prompt longer without making the goal clearer.
The Diagnostic Framework
When an AI response isn’t working, use the result itself to identify what may be missing.
| Symptom | Likely Missing Ingredient | Fix |
|---|---|---|
| The response addresses the topic but not the task | Instructions | Add a specific action, format, or length |
| The response is correct but generic | Context | Add relevant facts about your situation |
| The task and background are clear, but the tone or style is off | Examples | Add one sample showing the target style |
| The response covers too much or focuses on the wrong areas | Instructions | Narrow the scope |
| The response misses something you assumed was obvious | Context | State that information directly |
This is a starting point rather than a strict rule. Sometimes more than one ingredient is missing.
When that happens, fix the clearest problem first and try the prompt again before adding more.
A Common Misdiagnosis
One of the easiest mistakes to make is treating a context problem as an instruction problem.
When an answer feels generic, you may be tempted to add more formatting rules, rewrite the task, or make the instructions longer.
But if the task was already clear, that often won’t help.
The AI may simply be missing the information that makes the answer specific to your situation.
Before adding more instructions, ask:
Does the AI know enough about my situation to give me the answer I need?
If not, context is probably the better fix.
Practical Takeaway
When an AI response falls short, ask these three questions:
- Did it do the wrong thing?
Check the instructions. - Did it do the right thing, but give me a generic answer?
Add context. - Did it do the right thing, but in the wrong style?
Add an example.
That simple distinction can save you from repeatedly rewriting a prompt without knowing what needs to change.
Instead of adding more and more detail, give the AI the right kind of information.
FAQ
Do I need all three ingredients in every prompt?
No. Simple factual requests may only need a clear instruction. More complex or subjective tasks are more likely to benefit from context, examples, or both.
Which ingredient matters most?
Instructions are essential for defining the task. Context becomes especially important when the answer needs to reflect a particular business, audience, or situation. Examples are most useful when style, tone, or format matters.
Can too much context hurt a prompt?
Yes, if the extra information isn’t relevant. More detail doesn’t automatically make a prompt better. Include information that helps shape the answer and leave out details that don’t affect the result.
What if I’m not sure which ingredient is missing?
Start with the response you received. Identify what feels wrong, then use the diagnostic framework above to work backwards. It’s usually faster than changing several parts of the prompt at once.
