How to Use AI to Practise for Your Train Driver Interview

Why standard interview prep is not enough

Most candidates prepare for the train driver competency interview the wrong way. They read example questions, mentally rehearse what they would say, perhaps write a few notes, and arrive at the interview confident they have prepared. They then deliver an answer they have never actually said out loud in full, discover halfway through that the example they chose does not quite fit the question, and lose the assessor's attention.

The fix is to practise aloud, in full, against a specific question, and get feedback on what you actually said. This is exactly what AI tools are good at — and it is one of the few areas where using AI for job preparation makes a measurable difference rather than just a comforting one.

Using a generic chatbot (ChatGPT, Claude, Gemini)

You can absolutely use a general-purpose chatbot to practise. The major models are good at understanding the STAR framework, recognising weak answers, and giving polite, structured feedback. For very early-stage practice, this can be useful.

The drawbacks are real, though. A generic chatbot has no specific knowledge of what UK train driver assessors actually score. It does not know that safety questions carry disproportionate weight, that 'I would' instead of 'I did' is a deal-breaker, or that the rules-and-procedures questions are testing your willingness to follow procedures you disagree with. Its default register is encouraging, not honest. It is, in practice, too kind.

The other limitation is that you have to feed it the question, the assessment criteria, and the context yourself. Every time. The better the prompt you write, the better the feedback — but most candidates writing the prompt are exactly the people who do not yet know what good looks like.

Using a purpose-built tool

A purpose-built AI interview coach removes the prompt-engineering problem. The system prompt has already been written by someone who knows what UK train driver assessors are scoring against. The question bank is pre-loaded. The model is briefed to be honest, not polite, and to give the candidate specific things to add or change rather than vague encouragement.

We built the AI Interview Coach into Train Driver Tests for exactly this reason. The Coach is briefed as a fifteen-year train driver and assessor. It uses the same five competency areas UK train operators use — safety, rules and procedures, attention and focus, communication, motivation and resilience — and gives feedback addressed to you as a candidate. You speak or type your answer to any of the 20 prep questions, and within a few seconds you get a paragraph of feedback explaining what worked, what was weak, and what to add. The full feedback can be played back as audio if you want to listen rather than read.

Every account gets one Coach feedback free to try. Unlimited usage is a one-time £20 — no subscription, no expiry — because building real interview confidence takes more than a single attempt at a single question.

A practical workflow that actually works

Here is the routine that produces measurable improvement, regardless of which AI tool you use:

• ✓Pick one competency area at a time. Do not try to prepare all 20 questions in a single session. Start with safety — it is the highest-weighted area.
• ✓Read the question and the prompt for what assessors are looking for. Do not skip this step. You cannot answer a question well if you do not know what is being scored.
• ✓Brainstorm two specific real examples from your own past — work, voluntary, military, education — that fit. Pick the stronger one.
• ✓Speak the answer out loud, all the way through, using STAR. Do not stop to refine mid-answer. Get to the end. Time yourself — aim for ninety seconds.
• ✓Submit it to the AI. Read the feedback. Identify the single biggest weakness it flagged.
• ✓Re-answer the question, addressing only that weakness. Submit it again. Compare.
• ✓Move to the next question once you have two clean STAR answers in the same competency area.

What good AI feedback looks like

Good feedback identifies specifics. Bad feedback compliments your effort. If the AI is telling you 'great structure!' and 'nice example!', it is not helping you. The signs of useful feedback are different.

Useful feedback names the actual problem — 'you spent fifty seconds on the Situation and only fifteen on the Action; assessors score the Action.' It points to a specific phrase that hurts the answer — 'we communicated' should be 'I communicated to the signaller that...'. It tells you what to add — 'this example demonstrates teamwork but does not show safety instinct; assessors will mark you down on this question because they cannot see the safety angle.' If the tool is not doing this, switch to one that does.

Common weaknesses AI is good at spotting

Across hundreds of practice answers, certain patterns repeat. AI is reliably good at catching the following:

• ✓Answers that are too brief — under sixty words or roughly twenty seconds of speech. Real interview answers are 90–120 seconds.
• ✓'We' creep — answers that describe what a team did rather than your individual contribution.
• ✓Hypothetical framing — 'I would have...' instead of 'I did...'.
• ✓Missing safety angles — answers that demonstrate the competency but never mention safety, which is the highest-weighted criterion.
• ✓Generic phrasing — 'I love working with people' rather than 'I de-escalated a passenger dispute by...'.
• ✓Skipped Result — strong Situation and Action but no measured outcome at the end.

When AI is not enough

AI practice is excellent for content, structure, and breadth — getting you fluent across the question bank, exposing your weak examples, training you out of the 'we' habit. It is not a substitute for the human element of the actual interview. It cannot replicate a live assessor's probing follow-up questions, the pressure of being watched, or the social cost of going blank.

For that last mile, find a friend or family member to run a mock interview the day before. They do not need to know the rail industry — they just need to read a question from your list, listen quietly, and resist the urge to help. Combine months of AI practice with one or two real mocks in the final week and you will walk into the actual interview having said your answers out loud, in some form, dozens of times.