ISTQB CT-GenAI study plan

Explain how large language models generate output token by token, including the roles of training data, fine-tuning and inference parameters such as temperature and top-p.

Distinguish generative AI from classical ML and rule-based test tooling, naming two testing tasks each is and is not suited for.

Identify the main risks of using generative AI in testing (hallucination, data leakage, non-determinism, bias, IP and licensing) and map each to one mitigation.

Define prompt, context window, embedding and retrieval-augmented generation in your own words with a testing example for each.

Describe the ISTQB test process activities and state, for each, where generative AI can assist and where human judgement must remain.

Analyse three published or self-observed cases of generative AI defects in test artefacts and classify the root cause.

Draft a prompt template for requirements analysis that asks the model to list ambiguities, missing acceptance criteria and testability concerns.

Evaluate two prompting techniques (few-shot examples, role assignment, chain-of-thought) against a test-analysis task and record which produced more usable output.

Apply equivalence partitioning and boundary value analysis to a sample requirement, then use generative AI to cross-check the partition set and log every discrepancy.

Generate a risk-based test condition list from a user story and justify each prioritisation decision rather than accepting the model's ranking.

Derive test cases from a state transition or decision table model produced with AI assistance and verify the model against the source requirement.

Assess AI-generated test cases for coverage gaps, duplicates and untestable steps using a review checklist you build.

Convert approved test cases into Gherkin scenarios with generative AI and validate each scenario against the original acceptance criteria.

Produce test data sets (valid, invalid, boundary, privacy-sensitive) with AI assistance and check them for realism, coverage and personal data exposure.

Explain how retrieval-augmented generation over a project knowledge base changes prompt design compared with a bare chat prompt.

Identify where AI-generated test design output must be traced back to requirements for auditability.

Generate automated test scripts from Gherkin scenarios with an AI coding assistant and review the output for hard-coded values, brittle selectors and missing assertions.

Set up a small self-healing or selector-suggestion workflow and measure how many suggested fixes were correct.

Use generative AI to explain a failing test log and propose a root cause, then confirm or reject the hypothesis by manual investigation.

Describe how AI-assisted test generation fits into a CI pipeline, including gating, review and rollback.

Create synthetic test data and AI-generated exploratory test charters, then execute them against a sample application and record defects found.

Compare AI-generated test scripts against manually written equivalents on defect detection, maintainability and execution stability.

Apply a checklist to detect hallucinated APIs, non-existent assertions and silently passing tests in generated code.

Explain how to keep AI-generated tests deterministic and reproducible across runs.

Use generative AI to draft a defect report and a test summary report from raw execution data, then verify every figure against the source data.

Summarise a large set of test results into stakeholder-specific views and check the summary for unsupported claims.

Define metrics for evaluating generative AI use in testing, such as review rejection rate and defect escape rate.

Identify reporting risks including fabricated statistics, lost traceability and over-reliance on narrative summaries.

Build a continuous improvement loop that feeds review findings back into prompt templates and checklists.

Assess an organisation's readiness to adopt generative AI in testing against governance, data protection and skill criteria.

Consolidate weak areas from weeks 1 to 7 into a personal revision list and close each gap with a written explanation.

Complete a timed mock exam and analyse errors by domain to confirm balanced readiness.