AI Test Case Generation
Generate comprehensive test cases from natural language descriptions using AI.
AI Test Case Generation
The AI test case generation feature creates comprehensive, well-structured test cases from a natural language description of the feature or scenario you want to test. This is the fastest way to bootstrap a test suite for a new feature.
How to Generate Test Cases
- Navigate to the Test Cases tab in your project.
- Open the New dropdown in the toolbar and choose Create with AI.
- Choose where the requirement comes from:
- Paste Requirement — type or paste a description of the feature you want to test.
- Linked Requirement — pick a requirement already in the project and generate from it. Be as specific as possible. For example:
- "User login with email and password, including validation for empty fields, invalid email format, incorrect password, account lockout after 5 failed attempts, and successful login redirect to dashboard"
- "Shopping cart that supports adding items, updating quantities, removing items, applying discount codes, and calculating tax based on shipping address"
- Choose which Fields the AI should fill. Title is always generated; you can toggle Summary, Precondition, Expected Result, Test Data and Priority.
- Choose the Number of test cases to generate. This is required — pick a value between 1 and 10.
- Click Generate. The AI processes your description and produces test cases.
- Review the generated test cases in the preview. Each contains a title plus the fields you selected. Where the tester needs to perform a sequence of actions, those actions are written into the Summary as prose.
- Edit any test cases that need adjustment (modify titles, change priority, rewrite the summary).
- Tick the test cases you want to keep — Select All takes the lot — and click Submit (N) to add them to your project. Regenerate discards the batch and runs the generation again, which costs credits a second time.
AI generation does not populate the structured Test Steps table. Generated cases carry only the fields listed above; if you want discrete numbered steps with per-step expected results, add them by hand after saving.
Limits
| Limit | Value | What happens |
|---|---|---|
| Requirement length | 5,000 characters | The modal does not stop you pasting more, but generation fails with "Requirement must not exceed 5000 characters. Current length is N characters." Trim the spec to the part you want tested, or generate in sections. |
| Test cases per request | 1 to 10 | The field is required and defaults to 5. To produce more than 10, run the generation again — but see the note below on cost. |
| Response size | 16,384 tokens | The model's reply is capped, whether you asked for 1 test case or 10. If the reply is cut off mid-way the result cannot be read and generation fails with "Something went wrong while generating the AI response. Please try again." — ask for fewer cases, or untick fields you do not need. No credits are charged when this happens. |
| Credits | Token-based | Charged on the tokens sent and returned. Asking for 10 cases in one request is much cheaper than ten single-case requests, because your requirement and the instructions are sent once. See What an AI Action Costs. |
When your organization's credit balance reaches zero, generation is refused with a message that depends on the plan — a paid organization is told "You've used this month's AI credits. They refresh on the 1st, or add seats to raise your monthly allowance.", and a free organization "You've used your free AI credits. To use additional AI credits, upgrade to a Premium plan." If the owner has switched AI off entirely, it fails with "AI features are temporarily unavailable..." instead. See AI Privilege & Quality Settings.
Always review AI-generated test cases before saving. While the AI produces high-quality output, it may miss domain-specific edge cases or business rules that only your team knows.
Generation Tips
The quality of AI-generated test cases depends heavily on the quality of your input description. Here are tips to get the best results:
Write Specific Descriptions
| Instead of... | Try... |
|---|---|
| Test the login page | Test user login with email/password including empty field validation, invalid credentials, account lockout, remember me, and forgot password link |
| Test the checkout | Test e-commerce checkout flow: guest vs logged-in, credit card and PayPal, address validation, order summary, email confirmation |
| Test user settings | Test user profile settings: update display name, change email with verification, upload avatar (JPEG/PNG, max 5MB), change password requiring current password, delete account with confirmation |
Include Boundary Conditions
Mention edge cases, error states, and boundary conditions in your description. The AI will generate test cases covering these scenarios:
- Input limits (min/max values, character limits, file sizes)
- Error handling (network failures, server errors, invalid data)
- Permission scenarios (unauthorized access, expired sessions)
- Concurrent operations (multiple users, race conditions)
Specify the Testing Focus
If you want the AI to focus on specific areas, mention them explicitly:
- "Focus on negative testing" — generates more error and invalid input scenarios
- "Focus on security testing" — generates XSS, injection, and auth bypass scenarios
- "Focus on accessibility" — generates keyboard navigation, screen reader, and ARIA scenarios
- "Focus on performance" — generates load, timeout, and concurrency scenarios
You can generate test cases in batches. Start with a high-level description for broad coverage, then run additional generations focused on specific areas (security, edge cases, etc.) to deepen coverage.