All articles
Interview skills7 min read

How to Use AI for Interview Practice (Without Sounding Like a Robot)

By , Founder and Interview CoachLast updated

Judy M. Wallis is a former senior recruiter, an interview coach and the founder of Interview Practice Lab.

Share on LinkedIn

The paradox of AI interview preparation

Depending on the survey, anywhere from about a third to 80 percent of job seekers now use AI in their search (Gdoc, Clutch). They paste job descriptions into chatbots, ask for "perfect STAR answers", and read the resulting text off their screens.

Yet interview panels report a strange paradox: candidates have never sounded more rehearsed, and never less convincing.

Many recruiters and hiring managers now say they can hear what some call the "ChatGPT accent". It is smooth, generic sentences, an over-reliance on passive phrasing ("It was decided that a strategic initiative should be launched"), and corporate buzzwords ("synergised cross-functional deliverables") that no human would say across a boardroom desk.

When you sound like an algorithm, the panel tunes out. They cannot hear your genuine voice, your decision-making instincts, or the messy reality of how you actually solve workplace problems.

AI is one of the most powerful interview preparation tools ever created — but only if you use it as a sparring partner, not a teleprompter. Here is how to practice with AI so you walk in sounding confident, sharp and unmistakably human.

Gartner’s survey found 39% of candidates were already using AI in applications by late 2024. The useful distinction is not whether you use it, but whether it helps you prepare or substitutes for your own skills.

Rule 1: use AI to interrogate, never to script

The single biggest mistake candidates make is asking AI to write their answers for them: "Write a STAR answer for a time I dealt with a difficult stakeholder."

When AI writes your story from scratch, it invents generic conflict, inserts cliché resolutions, and strips out the small, authentic details that prove you did the work. Worse, you are now memorising someone else’s vocabulary. The moment nerves hit, your brain tries to recall the script word for word. If you lose your place, you freeze.

Instead, write your own rough draft first — even disjointed bullet points. Then feed it to the AI with this instruction:

"Act as a sceptical, senior hiring manager. Here is a rough draft of a story I plan to use for a stakeholder conflict question. Do not rewrite my story. Instead, poke holes in it. Tell me where I was too vague, where I took credit for team effort without explaining my own role, and what tough follow-up question you would ask me."

Now the AI is working as your interrogator. It forces you to think harder about real experience rather than handing you synthetic fluff.

Rule 2: break the "we" habit

Large language models are trained on billions of corporate web pages, annual reports and press releases. In that world, everything is written in the passive collective: "We identified the risk." "The team implemented the rollout."

In an interview room, speaking in "we" is lethal. Panels are assessing you, not your former colleagues. When every sentence begins with "we", interviewers cannot tell whether you led the project or sat at the back taking meeting minutes.

When reviewing your stories with AI, run a dedicated "I" audit:

  • Ask the tool: "Highlight every instance where I said we and suggest what specific individual action I need to clarify."
  • Replace "We overhauled the reporting process" with "I audited our three existing spreadsheet templates, found the team was duplicating five hours of data entry every Monday, and designed a single automated dashboard."

Rule 3: reading on screen is not spoken practice

Typing answers into a chat window uses visual and cognitive memory. Speaking aloud in front of three strangers under time pressure uses vocal cords, breath control and muscle memory. They are completely different pathways.

Almost every candidate has reviewed a well-crafted note on their laptop, felt completely prepared, then stumbled over their words within thirty seconds of the real interview.

  • Never read complete sentences. Reduce your notes to four or five anchor bullets — situation, task, action, result.
  • Practice answering out loud. Use voice-enabled coaching where you speak into your microphone, hear a realistic response, and have your pacing measured.
  • Listen for breath control and filler words. Notice where you get stuck moving from background into the action you took.

Before and after: the scripted trap

Here is what happens when you let AI write your answer, versus using AI to sharpen your own words. First, the AI-scripted version panels tune out:

"In my previous role as an operations lead, our organisation faced unprecedented challenges with supply chain bottlenecks. Recognising the imperative need for operational excellence, we spearheaded a strategic initiative to optimise vendor relationships. Through robust stakeholder alignment and agile communication protocols, we successfully mitigated delivery risks and delivered a seamless outcome for all parties involved."

The panel’s reaction: what did you actually do? What was the product? How long was the delay? This reads like an excerpt from an annual shareholder report.

Now the human version, refined through AI interrogation:

"Our main logistics partner unexpectedly increased delivery lead times from two days to nine, right before our peak quarter. I knew our frontline team could not absorb that without missing client deadlines. I pulled our last six months of dispatch data, identified our 20 highest-volume product lines, and called our account manager directly to negotiate a priority delivery window for those items. For everything else, I set up an automated daily dispatch tracker so our customer service reps could warn clients before problems landed. Our on-time delivery rate stayed above 94 percent through peak season, and we avoided an estimated $40,000 in expedited courier penalties."

The panel’s reaction: clear ownership, a realistic problem, concrete individual decisions and measurable commercial impact.

AI interview practice: a three-step routine

  • Jot the bones. Four bullets: what was the crisis, what was your specific responsibility, what two or three concrete actions did you take, what was the outcome.
  • Stress-test with prompts. Ask AI what is missing, to challenge your results, and to suggest three unexpected follow-up questions.
  • Rehearse aloud under time pressure. Aim for 60 to 90 seconds. Past two minutes means you spent too long setting up the situation and not enough explaining what you did.

If you want to hear how your spoken answers land right now, practice one question free with our interactive practice interview and get immediate feedback on your delivery, structure and pacing.

Where the line is: three zones

The test is whose thinking the panel is assessing. AI can help you prepare to show your skills; it should not replace them during the assessment.

  • Green: preparation, research and mock practice. Brainstorm examples, tighten your own draft, generate questions and get feedback on recorded practice answers. Keep the experience and decisions your own.
  • Amber: take-home tasks, cover letters, written applications and AI note-takers. Check the employer’s rules. An AI-written response with little of your input can sound generic or breach instructions. Ask if the rules are unclear.
  • Red: real-time tools feeding you answers during a live or recorded interview, AI-completed tests, or invented experience. Unless the employer explicitly authorises AI as part of the task, rely on your own thinking.

The ATO guidelines say candidates must rely on their own skills during the interview or assessment, should let the ATO know if they’ve used AI tools, and must explain how when asked. The APSC also publishes Principles for candidate use of AI in recruitment. Follow the instructions for your process.

Why employers are watching more closely

A Checkr survey found 59% of managers suspect candidates have used AI to misrepresent themselves. Countermeasures include camera pans, hand-wave tests, screen sharing, live writing and on-site final rounds.

A US survey by RefAssured found 74% of candidates had used AI to write or edit a resume, and 98% of HR professionals said they had caught candidates misrepresenting qualifications. Those figures describe different behaviours: using AI is not the same as misrepresenting yourself.

As Allwork.Space summed up employers’ view, AI can help candidates work, but it should not be able to impersonate their skills.

Can employers tell? Not with software, but answers give themselves away

Detection software is unreliable: OpenAI withdrew its classifier in 2023 for low accuracy. Panels often notice a different kind of evidence:

  • Polished but empty answers: impressive phrases without specific moments, names or numbers.
  • Long pauses and eyes moving off-screen before a suddenly fluent answer. These are not proof of AI use on their own.
  • An opener that collapses when the panel asks what you personally did next.
  • Written applications that do not match how you speak in the interview.

The risk is also sounding like everyone else. Answers shaped by the same tools converge on the same phrases. Specific, slightly imperfect examples from your own work are more useful than polished sameness.

What the rules say in Australia

From 10 December 2026, employers covered by the Privacy Act must describe automated or substantially computer-assisted decisions in their privacy policies, where those decisions could significantly affect you (OAIC guidance). This is a privacy-policy transparency rule, not an individual notice for each decision.

Covered employers already have APP 5 notice obligations about collecting and using your information. Most small businesses with turnover of $3 million or less are not covered, although exceptions apply. For recorded interviews, ask who reviews the recording and how long it is kept; Facecruit’s compliance guide discusses these questions.

There is no general duty for private employers to disclose AI use in recruitment, according to JobWatch. Anti-discrimination law still applies to AI screening. Read the privacy policy and ask how the process works.

This is general information, not legal advice. Employers set their own rules, so follow what the one you’re applying to has said.

Preparing for Australian panels

Plain, understated language is a better fit than US corporate jargon such as “ideating” or “thought partnership”. Explain what you delivered without inflation, and anchor each story to the capability framework and classification level in the job ad.

For APS 1–6, the Integrated Leadership System uses five names: Supports strategic direction; Achieves results; Supports productive working relationships; Displays personal drive and integrity; Communicates with influence. The wording changes at EL1 — for example, Shapes strategic thinking and Cultivates productive working relationships — as expectations broaden.

Ask AI to audit whether your own decisions demonstrate the advertised level rather than merely completing tasks. Queensland uses Leadership competencies for Queensland. The Selection Criteria & Government Applications Workbook provides frameworks and worked examples.

If they ask whether you used AI

Answer plainly and honestly: “I used AI to check structure and tighten wording; the examples and decisions are my own.” Explain what you actually did, and do not deny use you relied on.

If the interviewer itself is AI, read how to pass an AI interview screener. For recorded screens and Australian privacy rights, see AI and one-way video interviews in Australia.

Before the day, use the interactive AI interview checklist to check your preparation and setup.

Which AI can you practice with? (And what happened to Google Interview Warmup)

Google's free Interview Warmup tool has reportedly been retired (several sites date it to April 2026), and its old address now redirects to a general interview tips page. I can't find an explanation from Google. If you used it, or you're searching for something like it, the right replacement depends on what you need from your practice.

Free voice chat is the easiest place to start. ChatGPT's voice mode and Gemini Live, which Google now points people to for interview practice, let you talk through answers out loud at no cost. The catch is that the feedback is only as good as your prompt. Tell it the role, ask it to play a sceptical hiring manager, and ask it to challenge your answer rather than praise it. The prompt in Rule 1 above works well here.

A structured mock interview gives you something a chat doesn't: a consistent way to judge each answer. Our free practice interview asks questions aloud, listens to your spoken answers and shows you which part of the STAR structure is weakest. Paste in a job ad and it writes three tailored questions across three competencies; without one, you get standard questions across the same three. Each set includes at least one behavioural question drawn from your past experience and at least one situational question set in the role. You can try it with no sign-up.

The full Practice Lab goes further. It asks follow-up questions, mixes question types to suit the job ad you give it, and is designed to mimic a 1:1 mock interview session with a coach, so you practice the conversation rather than work through a list of questions.

For an interview that really matters, a real person is still the best test. AI is excellent for repetition, but a recruiter or coach can tell you how you came across: whether the story landed, whether you sounded confident, whether the panel would remember you. If the role matters a lot, book at least one 1:1 coaching session before the day.

Frequently asked questions

Is Google Interview Warmup still available?

No. Google appears to have retired it quietly: the old link now redirects to a general interview tips article, and several sites date the closure to April 2026. Google's own advice now points to Gemini Live for AI interview practice. You can also use ChatGPT's voice mode or our free AI interview coach.

Is it safe to use AI to prepare for an interview?

Yes, for preparation. Practicing your answers, asking for tough follow-up questions and getting feedback are all fair use. Two cautions: don't paste confidential information about a current or past employer into a chatbot, and don't use AI during a live interview unless the employer has said it's allowed. Policies are splitting, so read why some companies are banning AI in interviews and others are requiring it.

How often should I practice with AI?

Short and often beats one long session. As a starting point, do a few 15 to 20 minute sessions across the week before your interview, answering out loud in 60 to 90 seconds each time, then pick one thing to fix before the next.

Can AI replace a mock interview with a real person?

Not for an interview that matters. AI is excellent for repetition and for catching filler words or missing results. A person can tell you how you came across: your confidence, the energy in your story and whether a panel would remember it. Use AI for volume and a person for one final run-through.

What is the best free way to practice with AI?

Combine two things: free voice chat so you can talk out loud, and a structure such as STAR to judge your answers by. Our free practice interview does both in about a minute, with three questions, spoken answers and no sign-up.

Do I have to tell an employer I used AI?

Only if they ask, or their rules say so, and some public sector employers, such as the ATO, ask you to tell them up front. If they ask, be honest.

Can employers tell if I'm using AI in a video interview?

Increasingly, they try to. Long pauses, reading off-screen and answers that don't match your follow-up replies are the usual signs.

Can I use AI on a take-home task?

Check the instructions. If they don't say, ask.

Written by Judy M. Wallis, founder and interview coach at Interview Practice Lab and a former senior recruiter.