Every QA team has that one person now. In today’s AI in software testing landscape, it’s the one who pastes a feature description into ChatGPT, copies the output into Jira, and calls it done. Their test coverage looks great on paper. Their sprint metrics are green. And yet, the payment flow broke in production because nobody thought about what “test the checkout” actually means when your app supports three payment providers, two currencies, and a proration logic that kicks in mid-billing-cycle.
AI is a genuinely powerful tool for testers. But powerful tools misused just break things faster — and in QA, that means bugs in production.
According to the World Quality Report 2025–2026, published by Sogeti, Capgemini, and OpenText, the industry’s largest annual study on quality engineering, the adoption of Gen AI in testing has accelerated sharply, yet most organisations still lack a structured approach to using it effectively. That gap is where the problems live. This guide is about bridging it.
1. How to Write Better AI Prompts for Test Case Generation
AI doesn’t know your system. It knows patterns from everything it was trained on. Ask it something generic, and you get something generic back.
Bad prompt:
“Write test cases for a notifications feature.”
You’ll get the standard five: notification appears, notification is dismissed, notification redirects correctly, unread count updates, and maybe one about permission denied. Technically not wrong. Almost certainly incomplete.
Better prompt:
“Write test cases for an in-app notification system. Users receive notifications for: comment replies, status changes on test cases, and billing events. Notifications are real-time via WebSocket and are also sent as emails. Users can configure per-notification preferences. The app has both free and paid tiers. Paid users get email digests, free users don’t. Include edge cases for: WebSocket disconnection, notification preferences not saved, duplicate notifications, and notification behaviour when a user is logged in on two devices.”
The output is now actually useful. It accounts for the real states your system can be in, not just the happy path.
Think of AI like a smart new joiner on the team. The more context you give in the brief, the less time you spend fixing their work.
2. Use AI to Speed Up the Boring Stuff, Not Replace Thinking
AI is excellent at generating boilerplate. Test case scaffolding, repetitive boundary data, API request bodies, all of this can be drafted in seconds. The thinking part of what edge cases matter for this specific feature is still yours.
Live example:
You’re testing a file upload feature. Users can upload CSV reports up to 10MB. Generating 15 different test files by hand takes a good chunk of your morning.
Prompt AI:
“Generate 12 CSV test files for a report upload feature. Include: valid file under 10MB, file exactly at 10MB, file at 10.1MB, empty file, file with headers only and no data rows, file with missing required column ‘test_case_id’, file with duplicate row IDs, file with special characters in text fields, file with UTF-8 encoding, file with Windows line endings (CRLF), file with a mix of valid and invalid rows, and a file where numeric fields contain text values.”
Done in under a minute. You still need to verify the files match your actual validation logic, but the groundwork is laid. The time you saved goes into actually running the tests and observing behaviour, which is where testers add real value.
3. Watch Out for AI Hallucinations (This One Burns Teams)
This doesn’t get talked about enough in the QA community. AI confidently produces test assertions that are wrong. Not slightly off, fundamentally wrong. And because the output looks clean and professional, it slips through.
A real scenario:
You’re testing a subscription upgrade flow. A user moves from Free to Pro mid-cycle. You ask AI to help write assertions for the API response.
AI generates:
“Assert that the response returns HTTP 200 with { “status”: “upgraded”, “effective_date”: “<today>” }”
Your actual API returns HTTP 202 (accepted, async processing), and effective_date is the next billing cycle start date, not today. The test passes every single time. But the assertions were never validating the right thing.
AI doesn’t know your API contracts. It guesses based on common patterns it has seen before. This is especially dangerous in API testing, where wrong assertions create false confidence.
The fix is simple: Always cross-reference AI-generated assertions against your actual API documentation, OpenAPI specs, or a quick Postman run. Never assume the response structure it generates is correct for your system.
4. Using AI to Write Better Bug Reports
This is one of the most practical and underused applications of AI in QA. Most bug reports written in a hurry are vague, missing reproduction steps, or lack the context that developers actually need to investigate.
Take a rough note and let AI structure it into something that saves everyone time.
Before (what you typed at 5 PM on a Friday):
“Export button broken on the reports page, keeps loading, checked in Chrome.”
Prompt:
“Turn this into a structured bug report: Export button on reports page just spins and never downloads anything. Tested in Chrome. Steps: go to Reports > click Export CSV > button shows loading spinner > nothing downloads after 2 minutes. Expected: CSV file downloads. Actual: spinner keeps going, no download, no error message shown.”
After (AI output you edit and file):
Title: Export CSV button on Reports page hangs indefinitely, no file downloaded, and no error shown
Environment: Chrome 123, staging environment
Steps to Reproduce:
Navigate to the Reports page Click the “Export CSV” button Observe the loading spinner
Expected Result: CSV file downloads within a few seconds
Actual Result: Loading spinner continues indefinitely. No file is downloaded. No error message is displayed to the user.
Severity: Medium core reporting feature blocked for all users
The developer gets context they can actually use. You filed the bug in two minutes instead of ten. That’s the right use of AI.
5. Generating Gherkin Test Scenarios with AI
A lot of agile QA teams write acceptance tests in Gherkin (Given/When/Then format). AI is very good at this, but again, only if you tell it what your feature actually does.
Prompt:
“Write Gherkin scenarios for a team member invitation feature. A workspace admin can invite users by email. Invited users receive an email with an accept link valid for 48 hours. If the user already has an account, they are added directly. If they don’t, they go through a signup flow first. An admin can also resend or revoke pending invitations.”
AI Output (that you then review and adjust):
Scenario: Admin invites a new user successfully
Given I am logged in as a workspace admin
When I enter "newuser@example.com" in the invite field and click Send
Then the user should receive an invitation email
And the invitation status should show "Pending"
Scenario: Invited user with existing account accepts invitation
Given a user with "existing@example.com" already has an account
When they click the invite link in their email
Then they should be added to the workspace without going through the signup
Scenario: Invitation link expires after 48 hours
Given an invitation was sent 49 hours ago
When the invited user clicks the link
Then they should see an "Invitation expired" message
And they should be offered an option to request a new invite
Scenario: Admin revokes a pending invitation
Given there is a pending invitation for "pending@example.com"
When the admin clicks Revoke
Then the invitation should be removed from the pending list
And the invite link should no longer be valid
6. AI Tools Worth Knowing for Software Testers
Not all AI tools are the same. Here’s a practical breakdown of what actually works for different QA tasks:
Tool
Best For in QA
Watch Out For
ChatGPT / GPT-4o
Test case drafting, edge case brainstorming, bug report structuring
Generic output without detailed context
GitHub Copilot
AI pair programmer for writing, explaining, and refactoring all types of code, including test automation scripts in Playwright, Cypress, and Selenium
Outdated API usage in generated scripts
Claude
Analysing long PRDs, contracts, and large test plans
Less precise for code-specific tasks
Gemini
Google Workspace integration, quick lookups
Responses can be inconsistent with long context windows; less reliable for complex multi-step coding tasks compared to Copilot
Mabl / Testim
AI-powered end-to-end test maintenance
Needs initial setup investment
Applitools
Visual regression testing with AI comparison
Overkill for non-UI-heavy apps
7. AI in CI/CD Pipelines — Where It’s Actually Useful Now
AI is starting to show up not just in how testers write tests, but in how pipelines run and report them.
A few things are happening right now that are worth knowing:
Flaky test detection — Tools like Buildkite and some GitHub Actions integrations can now flag tests that pass and fail non-deterministically, cluster them, and suggest whether they’re environment issues or race conditions. This used to be a manual investigation.
Failure triage — Instead of parsing 300 lines of a failed build log, AI can summarise what broke, what the likely cause is, and which recent commits are connected. Teams using this are cutting investigation time significantly.
AI-assisted test selection — Rather than running the full regression suite on every PR, some pipelines now use AI to predict which tests are most relevant to the changed code. Faster feedback loops, same coverage confidence.
You don’t need to implement all of this tomorrow. But knowing it exists means you can have the right conversation with your engineering team when sprint planning comes around.
8. The 5 Most Common Mistakes Testers Make with AI
If you’re going to share one section from this blog, share this one.
Copying AI output without reading it. The output looks structured and professional, so it feels done. It’s not. Read every line before it goes into Jira or your test suite.
Using generic prompts for specific systems. “Write test cases for the dashboard” is not a prompt. It’s a request for guesswork. Describe your system, your users, and your edge cases.
Trusting AI with business logic. Prorated billing, custom permission models, multi-tenant data isolation — AI doesn’t know your rules. It’ll generate tests based on the most common version of similar features. Your feature might not be the most common version.
Using AI to skip exploratory testing. AI generates test cases based on requirements. It can’t catch the behaviour that only shows up when you actually use the product. Exploratory testing is irreplaceable.
Not iterating on the prompt. If the first output isn’t great, don’t give up — improve the context. Add constraints, add examples, tell it what you didn’t want. The second and third prompts are almost always better than the first.
9. What You Should Never Hand to AI
Some things should stay in the hands of the tester, full stop:
Go/no-go decisions before a release — AI can summarise test results. It cannot make a judgment call about business risk.
Exploratory testing — This requires curiosity, intuition, and knowing your users. No prompt captures that.
Security testing strategy — AI knows OWASP Top 10. It doesn’t know your threat model, your infrastructure, or which data is most sensitive in your specific context.
Risk-based test prioritisation — What to test most critically before a release is a judgement that requires understanding business impact, user behaviour, and technical complexity together. That’s a human job.
The Bottom Line
In AI in software testing, an average tester becomes faster at being average, while a sharp tester becomes significantly more productive. The difference comes down to whether you’re using AI to amplify your thinking or replace it.
The testers who look foolish with AI are the ones who copy-paste outputs without reading them, generate test cases without knowing their own system, and treat AI confidence as a substitute for correctness.
The testers who come out ahead are the ones who know exactly what question to ask, verify what comes back, and use the time saved to do the real work — understanding the system, talking to developers, and catching the classes of bugs that never show up in any template.
AI doesn’t know your app. You do. Keep it that way.
Tools Worth Bookmarking
ChatGPT— Test case drafting, bug report writing, Gherkin generation
GitHub Copilot — Automation scripting in Playwright / Cypress
Integrating Google Lighthouse with Playwright; Picture this: Your development team just shipped a major feature update. The code passed all functional tests. QA signed off. Everything looks perfect in staging. You hit deploy with confidence.
Then the complaints start rolling in.
“The page takes forever to load.” “Images are broken on mobile.” “My browser is lagging.”
Sound familiar? According to Google, 53% of mobile users abandon sites that take longer than 3 seconds to load. Yet most teams only discover performance issues after they’ve reached production, when the damage to user experience and brand reputation is already done.
The real problem isn’t that teams don’t care about performance. It’s that performance testing is often manual, inconsistent, and disconnected from the development workflow. Performance degradation is gradual. It sneaks up on you. And by the time you notice, you’re playing catch-up instead of staying ahead.
The Gap Between Awareness and Action
Most engineering teams know they should monitor web performance. They’ve heard about Core Web Vitals, Time to Interactive, and First Contentful Paint. They understand that performance impacts SEO rankings, conversion rates, and user satisfaction.
But knowing and doing are two different things.
The challenge lies in making performance testing continuous, automated, and actionable. Manual audits are time-consuming and prone to human error. They create bottlenecks in the release pipeline. What teams need is a way to bake performance testing directly into their automation frameworks to treat performance as a first-class citizen alongside functional testing.
Enter Google Lighthouse.
What Is Google Lighthouse?
Google Lighthouse is an open-source, automated tool designed to improve the quality of web pages. Originally developed by Google’s Chrome team, Lighthouse has become the industry standard for web performance auditing by Integrating Google Lighthouse with Playwright.
But here’s what makes Lighthouse truly powerful: it doesn’t just measure performance it provides actionable insights.
When you run a Lighthouse audit, you get comprehensive scores across five key categories:
Performance: Load times, rendering metrics, and resource optimization
Accessibility: ARIA attributes, color contrast, semantic HTML
Best Practices: Security, modern web standards, browser compatibility
SEO: Meta tags, mobile-friendliness, structured data
Progressive Web App: Service workers, offline functionality, installability
Each category receives a score from 0 to 100, with detailed breakdowns of what’s working and what needs improvement. The tool analyzes critical metrics like:
First Contentful Paint (FCP): When the first content renders
Largest Contentful Paint (LCP): When the main content is visible
Total Blocking Time (TBT): How long the page is unresponsive
Cumulative Layout Shift (CLS): Visual stability during load
Speed Index: How quickly content is visually populated
These metrics align directly with Google’s Core Web Vitals the signals that impact search rankings and user experience.
Why Performance Can’t Be an Afterthought
Let’s talk numbers, because performance isn’t just a technical concern it’s a business imperative.
Amazon found that every 100ms of latency cost them 1% in sales. Pinterest increased sign-ups by 15% after reducing perceived wait time by 40%. The BBC discovered they lost an additional 10% of users for every extra second their site took to load.
The data is clear: performance directly impacts your bottom line.
But beyond revenue, there’s the SEO factor. Since 2021, Google has used Core Web Vitals as ranking signals. Sites with poor performance scores get pushed down in search results. You could have the most comprehensive content in your niche, but if your LCP is above 4 seconds, you’re losing visibility.
The question isn’t whether performance matters. The question is: how do you ensure performance doesn’t degrade as your application evolves?
The Power of Integration: Lighthouse Meets Automation
This is where the magic happens when you integrate Google Lighthouse into your automation frameworks.
By Integrating Google Lighthouse with Playwright, Selenium, or Cypress, you transform performance from a periodic manual check into a continuous, automated quality gate.
Here’s what this integration delivers:
1. Consistency Across Environments
Automated Lighthouse tests run in controlled environments with consistent configurations, giving you reliable, comparable data across test runs.
2. Early Detection of Performance Regressions
Instead of discovering performance issues in production, you catch them during development. A developer adds a large unoptimized image? The Lighthouse test fails before the code merges.
3. Performance Budgets and Thresholds
You can set specific performance budgets for example, “Performance score must be above 90.” If a change violates these budgets, the build fails, just like a failing functional test.
4. Comprehensive Reporting
Lighthouse generates detailed HTML and JSON reports with visual breakdowns, diagnostic information, and specific recommendations. These reports become part of your test artifacts.
How Integration Works: A High-Level Flow
You don’t need to be a performance expert to integrate Lighthouse into your automation framework. The process is straightforward and fits naturally into existing testing workflows.
Step 1: Install Lighthouse Lighthouse is available as an npm package, making it easy to add to any Node.js-based automation project. It integrates seamlessly with popular frameworks.
Step 2: Configure Your Audits Define what you want to test which pages, which metrics, and what thresholds constitute a pass or fail. You can customize Lighthouse to focus on specific categories or run full audits across all five areas.
Step 3: Integrate with Your Test Suite Add Lighthouse audits to your existing test files. Your automation framework handles navigation and setup, then hands off to Lighthouse for the performance audit. The results come back as structured data you can assert against.
Step 4: Set Performance Budgets Define acceptable thresholds for key metrics. These become your quality gates if performance drops below the threshold, the test fails and the pipeline stops.
Step 5: Generate and Store Reports Configure Lighthouse to generate HTML and JSON reports. Store these as test artifacts in your CI/CD system, making them accessible for review and historical analysis.
Step 6: Integrate with CI/CD Run Lighthouse tests as part of your continuous integration pipeline. Every pull request, every deployment performance gets validated automatically.
The beauty of this approach is that it requires minimal changes to your existing workflow. You’re not replacing your automation framework you’re enhancing it with performance capabilities.
Practical Implementation: Code Examples
Let’s look at how this works in practice with a real Playwright automation framework. Here’s how you can create a reusable Lighthouse runner:
Feature: Integrating Google Lighthouse with the Test Automation Framework
This feature leverages Google Lighthouse to evaluate the performance,
accessibility, SEO, and best practices of web pages.
@test
Scenario: Validate the Lighthouse Performance Score for the Playwright Official Page
Given I navigate to the Playwright official website
When I initiate the Lighthouse audit
And I click on the "Get started" button
And I wait for the Lighthouse report to be generated
Then I generate the Lighthouse report
Decoding Lighthouse Reports: What the Data Tells You
Lighthouse reports are information-rich, but they’re designed to be actionable, not overwhelming. Let’s break down what you get:
The Performance Score
This is your headline number a weighted average of key performance metrics. A score of 90-100 is excellent, 50-89 needs improvement, and below 50 requires immediate attention.
Metric Breakdown
Each performance metric gets its own score and timing. You’ll see exactly how long FCP, LCP, TBT, CLS, and Speed Index took, color-coded to show if they’re in the green, orange, or red zone.
Opportunities
This section is gold. Lighthouse identifies specific optimizations that would improve performance, ranked by potential impact. “Eliminate render-blocking resources” might save 2.5 seconds. “Properly size images” could save 1.8 seconds. Each opportunity includes technical details and implementation guidance.
Diagnostics
These are additional insights that don’t directly impact the performance score but highlight areas for improvement things like excessive DOM size, unused JavaScript, or inefficient cache policies.
Passed Audits
Don’t ignore these! They show what you’re doing right, which is valuable for understanding your performance baseline and maintaining good practices.
Accessibility and SEO Insights
Beyond performance, you get actionable feedback on accessibility issues (missing alt text, poor color contrast) and SEO problems (missing meta descriptions, unreadable font sizes on mobile).
The JSON output is equally valuable for programmatic analysis. You can extract specific metrics, track them over time, and build custom dashboards or alerts based on performance trends.
Real-World Impact
Let’s look at practical scenarios where this integration delivers measurable value:
E-Commerce Platform
An online retailer integrated Lighthouse into their Playwright test suite, running audits on product pages and checkout flows. They set a performance budget requiring scores above 90. Within three months, they caught 14 performance regressions before production, including a third-party analytics script blocking rendering.
A B2B SaaS company added Lighthouse audits to their test suite, focusing on dashboard interfaces. They discovered their data visualization library was causing significant Total Blocking Time. The Lighthouse diagnostics pointed them to specific JavaScript bundles needing code-splitting.
Result: Reduced TBT by 60%, improving perceived responsiveness and reducing support tickets.
Content Publisher
A media company integrated Lighthouse into their deployment pipeline, auditing article pages with strict accessibility and SEO thresholds. This caught issues like missing alt text, poor heading hierarchy, and oversized media files.
Result: Improved SEO rankings, increased organic traffic by 23%, and ensured WCAG compliance.
The Competitive Advantage
Here’s what separates high-performing teams from the rest: they treat performance as a feature, not an afterthought.
By integrating Google Lighthouse with Playwright or any other automation framework, you’re building a culture of performance awareness. Developers get immediate feedback on the performance impact of their changes. Stakeholders get clear, visual reports demonstrating the business value of optimization work.
You shift from reactive firefighting to proactive prevention. Instead of scrambling to fix performance issues after users complain, you prevent them from ever reaching production.
Getting Started
You don’t need to overhaul your entire testing infrastructure. Start small:
Pick one critical user journey maybe your homepage or checkout flow
Add a single Lighthouse audit to your existing test suite
Set a baseline by running the audit and recording current scores
Define one performance budget perhaps a performance score above 80
Integrate it into your CI/CD pipeline so it runs automatically
From there, you can expand add more pages, tighten thresholds, incorporate additional metrics. The key is to start building that performance feedback loop.
Conclusion: Performance as a Continuous Practice
Integrating Google Lighthouse with Playwright; Web performance isn’t a one-time fix. It’s an ongoing commitment that requires visibility, consistency, and automation. Google Lighthouse provides the measurement and insights. Your automation framework provides the execution and integration. Together, they create a powerful system for maintaining and improving web performance at scale.
The teams that win in today’s digital landscape are those that make performance testing as routine as functional testing. They’re the ones catching regressions early, maintaining high standards, and delivering consistently fast experiences to their users.
The question is: will you be one of them?
Would you be ready to boost your web performance? You can start by integrating Google Lighthouse into your automation framework today. Your users and your bottom line will thank you.
Here’s a scenario that plays out in QA teams everywhere:
A tester spends 45 minutes manually writing test cases for a new feature. Another tester, working on the same type of feature, finishes in 12 minutes with better coverage, clearer scenarios, and more edge cases identified.
What’s the difference? Experience isn’t the deciding factor, and tools alone don’t explain it either. The real advantage comes from how they communicate with intelligent systems using effective QA Prompting Tips.
The testing world is changing more rapidly than we realise. Today, every QA engineer interacts with AI-powered tools, whether generating test cases, validating user stories, analysing logs, or debugging complex issues. But here’s the uncomfortable truth: most testers miss out on 80% of the value simply because they don’t know how to ask the right questions—especially when applying the right QA Prompting Tips.
That’s where prompting comes in.
Prompting isn’t about typing fancy commands or memorising templates. It’s about asking the right questions, in the right context, at the right time. It’s a skill that multiplies your testing expertise rather than replacing it.
Think of it this way: You wouldn’t write a bug report that just says “Login broken.” You’d provide steps to reproduce, expected vs. actual results, environment details, and severity. The same principle applies to prompting—specificity and structure determine quality, particularly when creating tests with QA Prompting Tips.
In this article, we’ll break down 10 simple yet powerful prompting secrets that can transform your day-to-day testing from reactive to strategic, from time-consuming to efficient, and from good to exceptional.
1. Context Is Everything
If you ask something vague, you’ll get vague answers. It’s that simple.
Consider these two prompts:
❌ Bad Prompt: “Write test cases for login.”
✅ Good Prompt: “You are a QA engineer for a healthcare application that handles sensitive patient data and must comply with HIPAA regulations. Write 10 test cases for the login module, focusing on data privacy, security vulnerabilities, session management, and multi-factor authentication.”
The difference? Context transforms generic output into actionable testing artifacts.
The first prompt might give you basic username/password validation scenarios. The second gives you security-focused test cases that consider regulatory compliance, session timeout scenarios, MFA edge cases, and data encryption validation, exactly what a healthcare app needs.
Why Context Matters
When you provide real-world details, AI tools can:
Align responses with your specific domain (fintech, healthcare, e-commerce)
Key Takeaway: Always include the “where” and “why” before the “what.” Context makes your prompts intelligent, not just informative, and serves as the foundation for effective QA Prompting Tips.
2. Define the Role Before the Task
Before you ask for anything, define what the system should think like. This single technique can elevate responses from junior-level to expert-level instantly.
✅ Effective Role Definition: “You are a senior QA engineer with 8 years of experience in exploratory testing and API validation. Review this user story and identify potential edge cases, security vulnerabilities, and performance bottlenecks.”
By assigning a role, you’re setting the expertise level, perspective, and focus area. The response shifts from surface-level observations to nuanced, experience-driven insights.
Role Examples for Different Testing Needs
For test case generation: “You are a detail-oriented QA analyst specializing in boundary value analysis…”
For bug analysis: “You are a senior test engineer experienced in root cause analysis…”
For automation: “You are a test automation architect with expertise in framework design…”
For performance: “You are a performance testing specialist, an expert in load testing methodologies and tools.”
Key Takeaway: Assign a role first, then give the task. It fundamentally changes the quality and depth of what you receive.
3. Structure the Output
QA engineers thrive on structured tables, columns, and clear formats. So ask for it explicitly.
✅ Structured Prompt: “Generate 10 test cases for the password reset feature in a table format with columns for: Test Case ID, Test Scenario, Pre-conditions, Test Steps, Expected Result, Actual Result, and Priority (High/Medium/Low).”
This gives you something that’s immediately copy-ready for Jira, TestRail, Zephyr, SpurQuality, or any test management tool. No reformatting. No cleanup. Just actionable test documentation.
Structure Options
Depending on your need, you can request:
Tables for test cases and test data
Numbered lists for test execution steps
Bullet points for quick scenario summaries
JSON/XML for API test data
Markdown for documentation
Gherkin syntax for BDD scenarios
Key Takeaway: Structured prompts produce structured results. Define the format, and you’ll save hours of manual reformatting.
4. Add Clear Boundaries
Boundaries create focus and prevent scope creep in your results.
✅ Bounded Prompt: “Generate exactly 8 test cases for the search functionality: 3 positive scenarios, 3 negative scenarios, and 2 edge cases. Focus only on the basic search feature, excluding advanced filters.”
This approach ensures you get:
The exact quantity you need (no overwhelming lists)
Scope: “Focus only on the checkout process, not the entire cart.”
Test types: “Only functional tests, no performance scenarios”
Priority: “High and medium priority only”
Platforms: “Web application only, exclude mobile”
Key Takeaway: Constraints keep your output precise, relevant, and actionable. They prevent information overload and maintain focus.
5. Build Step by Step (Prompt Chaining)
Just as QA processes are iterative, effective prompting follows a similar pattern. Instead of asking for everything at once, break it into logical steps.
Example Prompt Chain
Step 1:
“Analyze this user story and summarize the key functional requirements in 3-4 bullet points.”
Step 2:
“Based on those requirements, create 5 high-level test scenarios covering happy path, error handling, and edge cases.”
Step 3:
“Expand the second scenario into detailed test steps with expected results.”
Step 4:
“Identify potential automation candidates from these scenarios and explain why they’re suitable for automation.”
This layered approach produces clear, logical, and well-thought-out results. Each step builds on the previous one, creating a coherent testing strategy rather than disconnected outputs.
Key Takeaway: Prompt chaining mirrors your testing mindset. It’s iterative, logical, and produces higher-quality results than single-shot prompts.
6. Use Prompts for Reviews, Not Just Creation
Don’t limit AI tools to creation tasks; leverage them as your review partner.
Review Prompt Examples
✅ Test Case Review: “Review these 10 test cases for the payment gateway. Identify any missing scenarios, redundant steps, or unclear expected results.”
✅ Bug Report Quality Check: “Analyze this bug report and suggest improvements to make it clearer for developers. Focus on reproducibility, clarity, and completeness.”
✅ Test Summary Comparison: “Compare these two test execution summary reports and highlight which one communicates results more effectively to stakeholders.”
✅ Documentation Review: “Review this test plan and identify sections that lack clarity or need more detail.”
This transforms your workflow from one-directional (you create, you review) to collaborative (AI assists in both creation and quality assurance).
Key Takeaway: Use AI as your review partner, not just your assistant. It catches what you might miss and improves overall quality.
7. Use Real Scenarios and Data
Generic prompts produce generic results. Feed real test data, actual API responses, or specific scenarios for practical insights.
✅ Real-Data Prompt: “Here’s the actual API response from our login endpoint: {‘status’: 200, ‘token’: null, ‘message’: ‘Success’}. Even though the status is 200 and the message is success, this is causing authentication failures. What could be the root cause, and what test scenarios should I add to catch this in the future?”
This gives you:
Specific debugging insights based on actual data
Relevant test scenarios tied to real issues
Actionable recommendations, not theoretical advice
When to Use Real Data
Debugging: Paste actual logs, error messages, or API responses
Test data generation: Provide sample data formats
Scenario validation: Share actual user workflows
Regression analysis: Include historical bug patterns
Key Takeaway: Realistic inputs produce realistic testing insights. The more specific your input, the more valuable your output.
Note: Be cautious about the data you send to the AI model; it might be used for their training purpose. Always prefer a purchased subscription with a data privacy policy.
8. Set the Quality Bar
If you want a particular tone, standard, or level of professionalism, specify it upfront.
✅ Quality-Defined Prompts:
“Write concise, ISTQB-style test scenarios for the mobile registration flow using standard testing terminology.”
“Generate a bug report following IEEE 829 standards with proper severity classification and detailed reproduction steps.”
“Create BDD scenarios in Gherkin syntax following best practices for Given-When-Then structure.”
This instantly elevates the tone, structure, and professionalism of the output. You’re not getting casual descriptions, you’re getting industry-standard documentation.
Quality Standards to Reference
ISTQB for test case terminology
IEEE 829 for test documentation
Gherkin/BDD for behaviour-driven scenarios
ISO 25010 for quality characteristics
OWASP for security testing
Key Takeaway: Define the tone and quality standard upfront. It ensures outputs align with professional testing practices.
9. Refine and Iterate
Just like debugging, your first prompt won’t be perfect. And that’s okay.
After getting an initial result, refine it with follow-up prompts:
Initial Prompt: “Generate test cases for user registration.”
Refinement Prompts:
✅ “Add data validation scenarios for email format and password strength.”
✅ “Rank these test cases by priority based on business impact.”
✅ “Include estimated effort for each test case (Small/Medium/Large).”
✅ “Add a column for automation feasibility.”
Each iteration moves you from good to great. You’re sculpting the output to match your exact needs.
Iteration Strategies
Add missing elements: “Include security test scenarios”
Adjust scope: “Remove low-priority cases and add more edge cases”
Change format: “Convert this to Gherkin syntax”
Enhance detail: “Expand test steps with more specific actions”
Key Takeaway: Refinement is where you move from good to exceptional. Don’t settle for the first output iteration until it’s exactly what you need.
10. Ask for Prompt Feedback
Here’s a meta-technique: You can ask AI to improve your own prompts.
✅ Meta-Prompt Example: “Here’s the prompt I’m using to generate API test cases: [your prompt]. Analyze it and suggest how to make it more specific, QA-focused, and likely to produce better test scenarios.”
The system will reword, optimize, and enhance your prompt automatically. It’s like having a prompt coach.
What to Ask For
“How can I make this prompt more specific?”
“What context am I missing that would improve the output?”
“Rewrite this prompt to be more structured and clear.”
“What role definition would work best for this testing task?”
Key Takeaway: Always review and optimize your own prompts just like you’d review your test cases. Continuous improvement applies to prompting, too.
The QA Prompting Pyramid: A Framework for Mastery
Think of effective prompting as a pyramid. Each level builds on the previous one, creating a foundation for expert-level results.
Level
Principle
Focus
Impact
🧱 Base
Context
Relevance
Ensures outputs match your domain and needs
🎭 Level 2
Role Definition
Perspective
Elevates expertise level of responses
📋 Level 3
Structure
Clarity
Makes outputs immediately usable
🎯 Level 4
Constraints
Precision
Prevents scope creep and information overload
🪜 Level 5
Iteration
Refinement
Transforms good outputs into exceptional ones
🧠 Apex
Self-Improvement
Mastery
Continuously optimizes your prompting skills
Start at the base and work your way up. Master each level before moving to the next. By the time you reach the apex, prompting becomes second nature, a natural extension of your testing expertise.
Real-World Impact: How Prompting Transforms QA Work
Let’s look at practical scenarios where these techniques deliver measurable results:
Test Case Generation
A QA team at a fintech company used structured prompting to generate test cases for a new payment feature. By providing context (PCI-DSS compliance), defining roles (security-focused QA), and setting boundaries (20 test cases covering security, functionality, and edge cases), they reduced test case creation time from 3 hours to 25 minutes while improving coverage by 40%. This type of improvement becomes even more powerful when teams apply effective QA Prompting Tips in their workflows.
Bug Analysis and Root Cause Investigation
A tester struggling with an intermittent bug used real API response data in their prompt, asking for potential root causes and additional test scenarios. Within minutes, they identified a race condition that would have taken hours to debug manually.
Test Automation Strategy
An automation engineer used prompt chaining to develop a framework strategy starting with requirements analysis, moving to tool selection, then architecture design, and finally implementation priorities. The structured approach created a comprehensive automation roadmap in one afternoon.
Documentation Review
A QA lead used review prompts to analyze test plans before stakeholder presentations. The AI identified unclear sections, missing risk assessments, and inconsistent terminology issues that would have surfaced during the actual presentation.
The Competitive Advantage: Why This Matters Now
Here’s the reality: AI won’t replace testers, but testers who know how to prompt will replace those who don’t.
This isn’t about job security, it’s about effectiveness. The QA engineers who master prompting will:
Deliver faster without sacrificing quality
Think more strategically by offloading routine tasks
Catch more issues through comprehensive scenario generation
Communicate better with clearer documentation and reports
Stay relevant as testing evolves
Prompting is becoming as fundamental to QA as writing test cases or understanding requirements. It’s not a nice-to-have skill; it’s a must-have multiplier.
Getting Started: Your First Steps
You don’t need to master all 10 techniques overnight. Start small and build momentum:
First Week: Foundation
Practice adding context to every prompt
Define roles before tasks
Track the difference in output quality
Second Week: Structure
Request structured outputs (tables, lists)
Set clear boundaries on scope and quantity
Compare structured vs. unstructured results
Third Week: Advanced
Try prompt chaining for complex tasks
Use prompts for review and feedback
Experiment with real data and scenarios
Fourth Week: Mastery
Set quality standards in your prompts
Iterate and refine outputs
Ask for feedback on your own prompts
The key is consistency. Use these techniques daily, even for small tasks. Over time, they become instinctive.
Conclusion: Prompting as a Core QA Skill
Smart prompting is quickly becoming a core competency for QA professionals. It doesn’t replace your testing expertise; it multiplies it, especially when you use the right QA Prompting Tips.
When you apply these 10 techniques, you’ll notice how your test cases become more comprehensive, your bug reports clearer, your scenario planning sharper, and your overall productivity significantly higher. These improvements happen faster when you incorporate effective QA Prompting Tips into your daily workflow.
Remember this simple truth:
“The best testers aren’t those who work harder; they’re those who work smarter by asking better questions.”
So start today. Pick one or two of these techniques and apply them to your next testing task. Notice the difference. Refine your approach. And watch as your testing workflow transforms from reactive to strategic with the help of QA Prompting Tips.
The future of QA isn’t about replacing human intelligence with artificial intelligence. It’s about augmenting human expertise with intelligent tools, and prompting is the bridge between the two.
Your Next Steps
If you found these techniques valuable:
Share this article with your QA team and start a conversation about prompting best practices
Bookmark this guide and reference it when crafting your next prompt
Try one technique today, pick the easiest one, and apply it to your current task
Drop a comment below. What’s your go-to prompt that saves you time? What challenges do you face with prompting?
Follow for more. We’ll be publishing guides on advanced prompt patterns, AI-driven test automation, and QA productivity hacks
Your prompting journey starts with a single, well-crafted question. Make it count.
Automation always comes with surprises. Recently, I stumbled upon one such challenge while working on a scenario that required automating PDF download using Playwright to verify a PDF download functionality. Sounds straightforward, right? At first, I thought so too. But the web application I was dealing with had other plans.
The Unexpected Complexity
Instead of a simple file download, the application displayed the report PDF inside an iframe. Looking deeper, I noticed a blob source associated with the PDF. Initially, it felt promising—maybe I could just fetch the blob and save it. But soon, I realized the blob didn’t actually contain the full PDF file. It only represented the layout instructions, not the content itself.
Things got more interesting (and complicated) when I found out that the entire PDF was rendered inside a canvas. The content wasn’t static—it was dynamically displayed page by page. This meant I couldn’t directly extract or save the file from the DOM.
At this point, downloading the PDF programmatically felt like chasing shadows.
The Print Button Dilemma
To make matters trickier, the only straightforward option available on the page was the print button. Clicking it triggered the system’s file explorer dialog, asking me to manually pick a save location. While that works fine for an end-user, for automation purposes it was a dealbreaker.
I didn’t want my automation scripts to depend on manual interaction. The whole point of this exercise was to make the process seamless and repeatable.
Digging Deeper: A Breakthrough
After exploring multiple dead ends, I finally turned my focus back to Playwright itself. That’s when I discovered something powerful—Playwright’s built-in capability to generate PDFs directly from a page.
The key was:
Wait for the report to open in a new tab (triggered by the app after selecting “Print View”).
Bring this new page into focus and make sure all content was fully rendered.
Use Playwright’s page.pdf() function to export the page as a properly styled PDF file.
The Solution in Action
Here’s the snippet that solved it:
// Wait for new tab to open and capture it
const [newPage] = await Promise.all([
context.waitForEvent("page"),
event.Click("(//span[text()='OK'])[1]", page), // triggers tab open
]);
global.secondPage = newPage;
await global.secondPage.bringToFront();
await global.secondPage.waitForLoadState("domcontentloaded");
// Use screen media for styling
await global.secondPage.emulateMedia({ media: "screen" });
// Path where you want the file saved
const downloadDir = path.resolve(__dirname, "..", "Downloads", "Reports");
if (!fs.existsSync(downloadDir)) fs.mkdirSync(downloadDir, { recursive: true });
const filePath = path.join(downloadDir, "report.pdf");
// Save as PDF
await global.secondPage.pdf({
path: filePath,
format: "A4",
printBackground: true,
margin: {
top: "1cm",
bottom: "1cm",
left: "1cm",
right: "1cm",
},
});
console.log(`✅ PDF saved to: ${filePath}`);
Key Highlights of the Implementation
Capturing the New Tab The Print/PDF Report option opened the report in a new browser tab. Instead of losing control, we captured it with context.waitForEvent(“page”) and stored it in a global variable global.secondPage. This ensured smooth access to the report tab for further processing.
Switching to Print View The dropdown option was switched to Print View to ensure the PDF was generated in the correct layout before proceeding with export.
Emulating Screen Media To preserve the on-screen styling (instead of print-only styles), we used page.emulateMedia({ media: “screen” }). This allowed the generated PDF to look exactly like what users see in the browser.
Saving the PDF to a Custom Path A custom folder structure was created dynamically using Node.js path and fs modules. The PDFs were named systematically and stored under Downloads/ImageTrend/<date>/, ensuring organized storage.
Full-Page Export with Print Background Using Playwright’s page.pdf() method, we captured all pages of the report (not just the visible one), along with background colors and styles for accurate representation.
Clean Tab Management Once the PDF was saved, the secondary tab (global.secondPage) was closed, bringing the focus back to the original tab for processing the next incident report.
What I Learned
This challenge taught me something new: PDFs in web apps aren’t always what they seem. Sometimes they’re iframes, sometimes blob objects, and in trickier cases, dynamically rendered canvases. Trying to grab the raw file won’t always work.
But with Playwright, there’s a smarter way. By leveraging its ability to generate PDFs from a live-rendered page, I was able to bypass the iframe/blob/canvas complexity entirely and produce consistent, high-quality PDF files.
Conclusion:
What started as a simple “verify PDF download” task quickly turned into a tricky puzzle of iframes, blobs, and canvases. But the solution I found—automating PDF download using Playwright with its built-in PDF generation—was not just a fix, it was an eye-opener.
It reminded me once again that automation isn’t just about tools; it’s about understanding the problem deeply and then letting the tools do what they do best.
This was something new I learned, and I wanted to share it with all of you. Hopefully, it helps the next time you face a similar challenge.
Introduction to Cypress and TypeScript Automation:
Nowadays, the TypeScript programming language is becoming popular in the field of testing and test automation. Testers should know how to automate web applications using this new, trending programming language. Cypress and TypeScript automation can be integrated with Playwright and Cypress to enhance testing efficiency. In this blog, we are going to see how we can play with TypeScript and Cypress along with Cucumber for a BDD approach.
TypeScript’s strong typing and enhanced code quality address the issues of brittle tests and improve overall code maintainability. Cypress, with its real-time feedback, developer-friendly API, and robust testing capabilities, helps in creating reliable and efficient test suites for web applications.
Additionally, adopting a BDD approach with tools like Cucumber enhances collaboration between development, testing, and business teams by providing a common language for writing tests in a natural language format, making test scenarios more accessible and understandable by non-technical stakeholders.
In this blog, we will build a test automation framework from scratch, so even if you have never used Cypress, Typescript, or Cucumber, there are no issues. Together, we will learn from scratch, and in the end, I am sure you will be able to build your test automation framework.
Before we start building the framework and start with our discussion on the technology stack we are going to use, let’s first complete the environment setup we need for this project. Follow the steps below sequentially and let me know in the comments if you face any issues. Additionally, I am sharing the official website links just in case you want to take a look at the information on the tools we are using. Check here,
The first thing we need to make this framework work is Node.js, so ensure you have a node installed on the system. The very next thing to do is to have all the packages mentioned above installed on the system. How can you install them? Don’t worry; use the below commands.
So far, we have covered and installed all we need to make this automation work for us. Now, let’s move to the next step and understand the framework structure.
Framework Structure:
Let’s now understand some of the main players of this framework. As we are using the BDD approach assisted by the cucumber tool, the two most important players are the feature file and the step definition file. To make this more robust, flexible and reliable, we will include the page object model (POM). Let’s look at each file and its importance in the framework.
Feature File:
Feature files are an essential part of Behavior-Driven Development (BDD) frameworks like Cucumber. They describe the application’s expected behavior using a simple, human-readable format. These files serve as a bridge between business requirements and automation scripts, ensuring clear communication among developers, testers, and stakeholders.
Key Components of Feature Files
Feature Description:
A high-level summary of the functionality being tested.
Helps in understanding the purpose of the test.
Scenarios:
Each scenario represents a specific test case.
Follows a structured Given-When-Then format for clarity.
Scenario Outlines (Parameterized Tests):
Used when multiple test cases follow the same pattern but with different inputs.
Allows for better test coverage with minimal duplication.
Tags for Organization:
Tags like @smoke, @regression, or @critical help in organizing and running selective tests.
Makes it easier to filter and execute relevant scenarios.
Web App Automation Feature File:
Feature: Perform basic calculator operations
Background:
Given I visit calculator web page
@smoke
Scenario Outline: Verify the calculator operations for scientific calculator
When I click on number "<num1>"
And I click on operator "<Op>"
And I click on number "<num2>"
Then I see the result as "<res>"
Examples:
| num1 | Op | num2 | res |
| 6 | / | 2 | 3 |
| 3 | * | 2 | 6 |
@smoke1
Scenario: Verify the basic calculator operations with parameter
When I click on number "7"
And I click on operator "+"
And I click on number "5"
Then I see the result as "12"
API Automation Feature File:
Feature: API Feature
@api
Scenario: Verify the GET call for dummy website
When I send a 'GET' request to 'api/users?page=2' endpoint
Then I Verify that a 'GET' request to 'api/users?page=2' endpoint returns status
@api
Scenario: Verify the DELETE call for dummy website
When I send 'POST' request to endpoint 'api/users/2'
| name | job |
| morpheus | leader |
Then I verify the POST call
| req | endpoint | name | job | status |
| POST | api/users | morpheus | zion resident | 200 |
@api
Scenario: I send POST Request call and Verify the POST call Using Step Reusablity
When I send 'POST' request to endpoint 'api/users/2'
| req | endpoint | name | job |
| POST | api/users | morpheus | zion resident |
Then I verify the POST call
| req | endpoint | name | job | status |
| POST | api/users | morpheus | zion resident | 200 |
Step Definition File:
Step definition files act as the implementation layer for feature files. They contain the actual automation logic that executes each step in a scenario. These files ensure that feature files remain human-readable while the automation logic is managed separately.
Key Components of Step Definition Files
Mapping Steps to Code:
Each Given, When, and Then step in a feature file is linked to a function in the step definition file.
Ensures test steps execute the corresponding automation actions.
Reusability and Modularity:
Common steps can be reused across multiple scenarios.
Avoid duplication and improve maintainability.
Data Handling:
Step definitions can take parameters from feature files to execute dynamic tests.
Enhances flexibility and test coverage.
Error Handling & Assertions:
Verifies expected outcomes and reports failures accurately.
Helps in debugging test failures efficiently.
Web App Step Definition File:
import { When, Then, Given } from '@badeball/cypress-cucumber-preprocessor'
import { CalPage } from '../../../page-objects/CalPage'
const calPage = new CalPage()
Given('I visit calculator web page', () => {
calPage.visitCalPage()
cy.wait(6000)
})
Then('I see the result as {string}', (result) => {
calPage.getCalculationResult(result)
calPage.scrollToHeader()
})
When('I click on number {string}', (num1) => {
calPage.clickOnNumber(num1)
calPage.scrollToHeader()
})
When('I click on operator {string}', (Op) => {
calPage.clickOnOperator(Op)
calPage.scrollToHeader()
})
API Step Definition File:
import { Given, When, Then } from '@badeball/cypress-cucumber-preprocessor'
import { APIUtility } from '../../../../Utility/APIUtility'
const apiPage = new APIUtility()
When('I send a {string} request to {string} endpoint', (req, endpoint) => {
apiPage.getQuery(req, endpoint)
})
Then(
'I Verify that a {string} request to {string} endpoint returns status',
(req, endpoint) => {
apiPage.iVerifyGETRequest(req, endpoint)
},
)
Then('I verify that {string} request to {string} endpoint', (datatable) =>
apiPage.postQueryCreate(datatable)
})
Then('I verify the POST call', (datatable) => {
apiPage.postQueryCreate(datatable)
})
When('I send {string} request to endpoint {string}', (req, endpoint) => {
apiPage.delQueryReq(req, endpoint)
})
Then(
'I verify {string} request to endpoint {string} returns status',
(req, endpoint) => {
apiPage.delQueryReq(req, endpoint)
},
)
Page File:
Page files in test automation frameworks serve as a structured way to interact with web pages while keeping test scripts clean and maintainable. These files typically encapsulate locators and actions related to a specific page or component within the application under test.
Key Components of Page Files in Test Automation Frameworks
Navigation Methods:
Functions to visit the required page using a URL or base configuration.
Ensures tests always start from the correct application state.
Element Interaction Methods:
Functions to interact with buttons, input fields, dropdowns, and other UI elements.
Encapsulates actions like clicking, typing, or selecting options to maintain reusability.
Assertions and Validations:
Methods to verify expected outcomes, such as checking if an element is visible or a value is displayed correctly.
Helps in ensuring the application behaves as expected.
Reusability and Modularity:
Each function is designed to be reusable across multiple test cases.
Keeps automation scripts clean by avoiding redundant code.
Handling Dynamic Elements:
Includes waits, scrolling, or retries to ensure elements are available before interaction.
Reduces flakiness in tests.
Test Data Handling:
Functions to pass dynamic test data and execute actions accordingly.
API utility files are essential in automated testing as they provide reusable methods to interact with APIs. These files help testers perform API requests, validate responses, and maintain structured automation scripts.
By centralizing API interactions in a dedicated utility, we can improve test maintainability, reduce duplication, and ensure consistent validation of API responses.
Key Components of an API Utility File:
Making API Requests Efficiently:
Functions for sending GET, POST, PUT, and DELETE requests.
Uses dynamic parameters to handle different endpoints and request types.
Response Validation & Assertions:
Ensures correct HTTP status codes are returned.
Validates response bodies for expected data formats.
Logging & Debugging:
Captures API request and response details for debugging.
Provides meaningful logs to assist in troubleshooting failures.
Handling Dynamic Data:
Supports dynamic payloads using external test data sources.
Allows testing multiple scenarios without modifying the core test script.
Error Handling & Retry Mechanism:
Implements error handling to manage unexpected API failures.
Can include automatic retries for transient errors (e.g., 429 rate limiting).
Security & Authentication Handling:
Supports authentication headers (e.g., tokens, API keys).
Ensures tests adhere to security best practices like encrypting sensitive data.
Currently, the base URL is fetched from Cypress.env(‘api_URL’), but we can extend it to support multiple environments (e.g., dev, staging, prod).
Enhance Error Handling & Retry Logic:
Implement a retry mechanism for APIs that occasionally fail due to network issues.
Improve error messages by logging API response details when failures occur.
Support Query Parameters & Headers:
Modify functions to accept optional query parameters and custom headers for better flexibility.
Improve Response Validation:
Extend validation beyond just checking the status code (e.g., validating response schema using JSON schema validation).
Use Utility Functions for Reusability:
Extract common assertions (e.g., checking response status, verifying keys in the response) into separate utility functions to avoid redundancy.
Implement Rate Limiting Controls:
Introduce a delay between API requests in case of rate-limited endpoints to prevent hitting request limits.
Better Logging & Reporting:
Enhance logging to provide detailed information about API requests and responses.
Integrate with test reporting tools to generate detailed API test reports.
Configuration Files:
Cypress.config.ts:
The Cypress configuration file (cypress.config.ts) is essential for defining the setup, plugins, and global settings for test execution. It helps in configuring test execution parameters, setting up plugins, and customizing Cypress behavior to suit the project’s needs.
This file ensures that Cypress is properly integrated with necessary preprocessor plugins (like Cucumber and Allure) while defining critical environment variables and paths.
Key Components of the Configuration File:
Importing Required Modules & Plugins:
Cypress needs additional plugins for Cucumber support and reporting.
@badeball/cypress-cucumber-preprocessor is used for running .feature files with Gherkin syntax.
@shelex/cypress-allure-plugin/writer helps in generating test execution reports using Allure.
@esbuild-plugins/node-modules-polyfill ensures compatibility with Node.js modules.
Setting Up Event Listeners & Preprocessors:
The setupNodeEvents function is responsible for handling plugins and configuring Cypress behavior dynamically.
The Cucumber preprocessor generates JSON reports and processes Gherkin-based test cases.
Browserify is used as the file preprocessor, allowing TypeScript support in tests.
Environment Variables & Custom Configurations:
api_URL: Stores the base API URL used for API testing.
screenshotsFolder: Defines the folder where Cypress will save screenshots in case of failures.
Defining E2E Testing Behavior:
setupNodeEvents: Attaches the preprocessor and other event listeners.
excludeSpecPattern: Ensures Cypress does not pick unwanted file types (*.js, *.md, *.ts).
specPattern: Specifies that Cypress should look for .feature files in cypress/e2e/.
baseUrl: Defines the website URL where tests will be executed (https://www.calculator.net/).
import { defineConfig } from 'cypress'
import { addCucumberPreprocessorPlugin } from '@badeball/cypress-cucumber-preprocessor'
import browserify from '@badeball/cypress-cucumber-preprocessor/browserify'
import allureWriter from '@shelex/cypress-allure-plugin/writer'
const {
NodeModulesPolyfillPlugin,
} = require('@esbuild-plugins/node-modules-polyfill')
async function setupNodeEvents(
on: Cypress.PluginEvents,
config: Cypress.PluginConfigOptions,
): Promise<Cypress.PluginConfigOptions> {
// This is required for the preprocessor to be able to generate JSON reports after each run, and more,
await addCucumberPreprocessorPlugin(on, config)
allureWriter(on, config),
on(
'file:preprocessor',
browserify(config, {
typescript: require.resolve('typescript'),
}),
)
// Make sure to return the config object as it might have been modified by the plugin.
return config
}
export default defineConfig({
env: {
api_URL: 'https://reqres.in/',
screenshotsFolder: 'cypress/screenshots',
},
e2e: {
// We've imported your old cypress plugins here.
// You may want to clean this up later by importing these.
setupNodeEvents,
excludeSpecPattern: ['*.js', '*.md', '*.ts'],
specPattern: 'cypress/e2e/**/*.feature',
baseUrl: 'https://www.calculator.net/',
},
})
Tsconfig.json:
The tsconfig.json file is a TypeScript configuration file that defines how TypeScript code is compiled and interpreted in a Cypress test automation framework. It ensures that Cypress and Node.js types are correctly recognized, allowing TypeScript-based test scripts to function smoothly.
Key Components oftsconfig.json:
compilerOptions (Compiler Settings)
“esModuleInterop”: true
Allows interoperability between ES6 modules and CommonJS modules, enabling seamless imports.
“target”: “es5”
Specifies that the compiled JavaScript should be compatible with ECMAScript 5 (older browsers and environments).
“lib”: [“es5”, “dom”]
Includes support for ES5 and browser-specific APIs (DOM), ensuring compatibility with Cypress test scripts.
“types”: [“cypress”, “node”]
Adds TypeScript definitions for Cypress and Node.js, preventing type errors in test scripts.
include (Files Included for Compilation)
**/*.ts
Ensures that all TypeScript files in the project directory are included in compilation.
The package.json file is a key component of a Cypress-based test automation framework that defines project metadata, dependencies, scripts, and configurations. It helps manage all the required libraries and tools needed for running, reporting, and processing test cases efficiently.
Key Components of package.json:
Project Metadata
“name”: “spurtype” → Defines the project name.
“version”: “1.0.0” → Specifies the current project version.
“description”: “Cypress With TypeScript” → Describes the purpose of the project.
Scripts (Commands for Running Tests & Reports)
“scr”: “node cucumber-html-report.js”
Runs a script to generate a Cucumber HTML report.
“coms”: “cucumber-json-formatter –help”
Displays help information for Cucumber JSON formatter.
“api”: “./node_modules/.bin/cypress-tags run -e TAGS=@api”
Executes Cypress tests tagged as API tests (@api).
“smoke”: “./node_modules/.bin/cypress-tags run -e TAGS=@smoke”
Executes smoke tests (@smoke) using Cypress.
“smoke4”: “cypress run –env allure=true,TAGS=@smoke1”
Runs a specific set of smoke tests (@smoke1) while enabling Allure reporting.
This script generates a Cucumber HTML report from JSON test results using the multiple-cucumber-html-reporter package. It extracts test execution details, including browser, platform, and environment metadata, and saves the output as an HTML file for easy visualization of test results in Cypress and TypeScript Automation.
The script requires the package to process JSON reports and generate an interactive HTML report.
Configuration Options
jsonDir → Specifies the location of Cucumber-generated JSON reports.
reportPath → Sets the directory where the HTML report will be saved.
reportName → Defines a custom name for the report file.
pageTitle → Sets the title of the generated HTML report page.
displayDuration → Enables duration display for each test case execution.
openReportInBrowser → Automatically opens the HTML report after generation.
Metadata Section
Browser: Specifies the test execution browser and version.
Device: Identifies the test execution machine.
Platform: Defines the operating system used for testing.
Custom Data Section
Provides additional test details such as Project Name, Test Environment, Execution Time, and Tester Information.
Cypress-cucumber-preprocessor.json
This JSON configuration file is primarily used to manage the Cypress Cucumber preprocessor settings. It enables JSON logging, message output, and HTML report generation, and it specifies the location of step definition files.
Specifies the directory where step definition files are located. These files contain the implementation for Gherkin feature file steps.
Conclusion:
Cypress and TypeScript together create a powerful and efficient framework for both web applications and API automation. By leveraging Cypress’s fast execution and robust automation capabilities alongside TypeScript’s strong typing and code scalability, we can build reliable, maintainable, and scalable test suites.
With features like Cucumber BDD integration, JSON reporting, HTML test reports, and API automation utilities, Cypress enables seamless test execution, while TypeScript enhances code quality, error handling, and developer productivity. The structured approach of defining page objects, API utilities, and configuration files ensures a well-organized framework that is both flexible and efficient.
As automation testing continues to evolve, integrating Cypress with TypeScript proves to be a future-ready solution for modern software testing needs. Whether it’s UI automation, API validation, or end-to-end testing, this dynamic combination offers speed, accuracy, and maintainability, making it an essential choice for testing high-quality web applications.