Stop 13 · AI & Modern QA PracticesTopics · 5/7

Prompt Engineering for QA

In short

Ask an AI 'write tests for login' and you'll get generic fluff. Give it the role, the rules and the format you want, and it becomes a genuinely useful testing partner.

Writing clear, structured prompts to get useful, testing-relevant output from an AI tool — providing context (requirements, existing test patterns) rather than a single vague sentence.

The purple blob explains an order with hand gestures at a chai stall while the chaiwala pours tea.

Think of it like this

Ordering at the chai stall

Say 'one chai' and you get whatever the chaiwala usually makes. Say 'one cutting, less sugar, extra adrak, no malai' and you get exactly your chai. A good prompt is a precise order, not a vague wish.

Watch it happen

Building a testing prompt, piece by piece

Role

Context

Task

Rules

Format

1/5Tell it who to be: a senior QA engineer testing a food delivery app.

The big ideas

What you need to know

Context beats cleverness

The biggest improvement comes from pasting the real requirement, not from magic phrases.

Show an example

Paste one test case in your team's format. The tool will copy the style for the rest.

Iterate

Treat it as a conversation: 'add negative cases for payment failure', 'remove duplicates', 'explain case 7'.

Safe inputs only

Use approved tools and remove passwords, customer data and secrets before pasting anything.

See it in code

Walk through the code

text

1You are a senior QA engineer testing a food delivery app.2 3Context: the user story and rules are below.4Story: As a customer, I can apply one coupon per order.5Rules: minimum order 500 rupees, coupon expires 31 Dec, one use per user.6 7Task: write test cases for the coupon feature.8Include positive, negative and boundary cases.9Do not invent rules that are not listed. Ask if anything is unclear.10 11Format: a table with ID, Title, Steps, Test data, Expected result.

1/4The role sets the tone and level of detail.

A reusable prompt template. Swap the story and rules for each feature and keep the rest.

Save yourself the pain

Common mistakes

Oops

Writing one vague sentence and blaming the tool for generic results.

Do this instead

Add role, context, rules and format. Most bad output comes from thin input.

Oops

Trusting expected results the tool calculated, like discount totals.

Do this instead

Check every number yourself. AI tools can make arithmetic and logic mistakes confidently.

Your mission

Vague vs structured

~20 min

Quick check

Which change usually improves AI-generated test cases the most?

Go deeper

1Practice iterating on a testing-related prompt 3-4 times to see how wording changes shift output.
2Learn to provide example test cases in a prompt for more consistent AI-generated output.
3Build a small library of prompt templates for recurring QA tasks (test case drafts, bug report drafts).