Testing with AI, and testing the AI itself
Two different jobs wear the same name. One is using a model to get your own work done faster. The other is testing a system whose output is different every time you run it – where “expected result” stops being a single value. This page keeps them apart and gives you the material for both.
Using AI to do testing work
The part you can start today, on work you already have.
Prompts that earn their place
Generating test cases from a requirement, turning a bug into a report someone can act on, reading an unfamiliar codebase, and writing the first draft of a script. With the prompts, not just the idea of them.
The tools, on the site
The AI test case prompt generator sits with the rest of the QA tools, next to the JSON, regex and XPath helpers you already use.
Testing a system that answers differently every time
This is the part interviews are starting to ask about, and the part most testers have never had to think through.
- What an assertion means when the output is not fixed
- Grounding and hallucination – how you prove one happened
- Evaluation sets, scoring and why a pass rate replaces pass/fail
- Prompt injection and the things a model should refuse
- Cost, latency and token limits as test conditions
- Regression when the model underneath you changes
The course
Ten modules, from the maths and Python up through machine learning, deep learning, generative AI, RAG, agents and then evaluation and deployment – which is where the testing lives.
Projects and interviews
AI projects
A voice-controlled assistant, blockchain certificate verification and an AI test case generator – each with the testing story that makes it worth putting on a CV rather than just building. The code is on the Download Center.
AI testing interview rounds
Basic, intermediate, advanced and scenario-based, written for the questions that are actually being asked now rather than the ones that sound impressive.