ISTQB CT-GenAI
GenAI in Regression Testing and Test Monitoring
Easy
Regression testing with GenAI
- Why are regression tests particularly well suited to automation?
- In which environment do automated regression tests typically run particularly often?
- In keyword-driven test automation, what does a keyword stand for?
- What is the purpose of a GenAI-supported impact analysis in regression testing?
- What is meant by self-healing tests?
- What can GenAI compile automatically for defect reports in regression testing?
- What applies to test scripts and reports that GenAI has produced in regression testing?
- Why do automated GUI tests often become unstable?
- Which change typically causes trouble for API regression tests?
Test monitoring and test control
- Where is the data for test monitoring often already available for GenAI to evaluate?
- What can GenAI achieve in test monitoring beyond plain numbers?
- How can GenAI support test control?
- What can GenAI contribute to test completion?
- What is the advantage of natural language summaries alongside a dashboard?
- Which metric is a typical result when an LLM prepares test data for monitoring?
Choosing a prompting technique
- For which tasks is prompt chaining particularly suitable?
- When is few-shot prompting the obvious choice?
- What is the core of meta prompting?
- May several prompting techniques be combined for a single test task?
- A team wants 40 test cases written in Gherkin format, always with the same structure. Which technique fits best?
Medium
Regression testing with GenAI
- A team has a keyword library for a lending app. How does an LLM help most when generating new regression scripts?
- Which prompting technique is well suited to having keyword-driven test scripts generated in a uniform form?
- In a bike rental app, only the price calculation for weekend rentals was changed. What does an AI-supported impact analysis provide?
- In a pharmacy web shop, the field "ZIP" is now called "Postal code"; the function is unchanged. What does a self-healing test do?
- What long-term goal do self-healing tests pursue?
- An LLM produces a report after each nightly regression run. What sensibly belongs in it?
- A regression run yields 30 failures. Which first step of a structured report analysis with an LLM makes sense?
- What is the purpose of a list of known anomalies in AI-supported analysis of regression results?
- A rental API will require a new authentication method. How can GenAI support the API regression tests?
- What is meant by dynamic locators to which GenAI can adapt GUI scripts?
- What forward-looking information can an AI-supported test report provide in regression testing?
Test monitoring and test control
- An LLM receives an export from the test management tool of a hospital app. Which result is most useful for test monitoring?
- Monitoring shows that tests for the medication list are far behind plan. Which GenAI help fits test control?
- What benefit do GenAI-created completion reports have beyond the individual project?
- What characterizes a dynamic dashboard that GenAI helps to build?
- An LLM has calculated metrics for the dashboard of a daycare management app from a tool export. What should happen before publication?
Choosing a prompting technique
- From user stories of a railway app, test conditions, then test cases and finally scripts are to be produced; each stage is to be checked. Which technique fits?
- A team wants log files examined for anomalies for the first time and has no suitable prompt yet. Which technique is a good start?
- Which sequence describes a sensible combination of the three prompting techniques?
- For which test task is meta prompting typically well suited?
Hard
Regression testing with GenAI
- After a change to the discount logic of a ticket shop, an AI impact analysis recommends only 12 of 400 regression tests. How does a test team best proceed?
- A self-healing test adapts because the "Order" button has disappeared and now clicks "Save for later". What does this case show?
- An LLM generates scripts for a keyword-driven framework, but some abort because keywords are misspelled. Which measure helps most specifically?
- A team sets up an AI assistant via a system prompt that checks and corrects keyword-driven scripts. Which detail most belongs in this system prompt?
- A structured analysis of regression results runs in a single LLM conversation. Which order makes most sense?
- An LLM groups 25 failed tests of an insurance app into three defect clusters. What risk exists here?
- For which types of regression tests can GenAI be used?
- A parcel tracking API now returns a "tracking" object with subfields instead of "status". What is important when adapting the tests with GenAI?
- Why does GenAI fit well with maintaining regression suites in fast release cycles?
- An AI-supported impact analysis is to select regression tests in a CI pipeline. Which prerequisite is most important for this?
Test monitoring and test control
- Based on the test data, an LLM reports that the defect rate in the "Billing" module is rising sharply. What is the best next step?
- Two weeks before release, an LLM suggests postponing low-risk tests to test critical areas more thoroughly. How should this be classified?
- What distinguishes test control from test monitoring when both are supported by GenAI?
- An LLM writes a summary of the test status for senior management. What is the greatest danger?
- Test data from several tools is partly unstructured, such as free-text comments. What is a typical contribution of GenAI to test monitoring?
Choosing a prompting technique
- A team wants weekly regression reports always in the same table format. Why is few-shot prompting superior to meta prompting here?
- Test scripts for a banking app are to be built step by step; each test step is to be checked for functional correctness. What speaks against one single large prompt?
- What risk exists if a prompt generated by meta prompting is used unchecked?
- Which property of a test task speaks most strongly for prompt chaining rather than few-shot prompting?
- In a prompt chain for a travel expense app, step 2 delivers test cases with wrong expected results. How do you proceed?