Can AI run a mock interview with me?
Yes. Both ChatGPT and Gemini have voice modes that let you speak your answers out loud, get spoken follow-up questions back, and read a transcript afterwards, so an AI can run a real mock interview rather than just hand you a list of ChatGPT interview questions. For AI interview practice that covers the part a written list can't: saying an answer out loud, in order, under time pressure. It does not replace the judgement of a person who knows the job, and every factual thing it produces still needs checking.
Tools that work
ChatGPT (voice inside a chat)
ChatGPT Voice lets you speak with ChatGPT and hear a spoken response, within a chat, so you can also read the response as text. The "Live" option can listen and speak at the same time, so you can interrupt or keep talking while it responds — that turn-taking is what makes it feel like an interview instead of a form.
Which voice options you see depends on your plan, workspace settings, region, app version and parental controls, and Live usage in Chat mode is measured over a rolling 24-hour period with plan limits. Audio clips from Live and Advanced Voice conversations are stored with the chat transcript and retained for 30 days, so think before you describe anything confidential about a current employer.
On plans: the ChatGPT pricing page lists a Free plan with unlimited text chats but limited messages with uploads, limited image creation, limited voice, limited deep research and limited memory; paid individual tiers shown are Go, Plus and Pro. Plus is $20 per month billed monthly, and OpenAI notes there is no annual billing option for Plus. Go is the low-cost paid tier with higher limits than Free for image generation, file uploads and data analysis — the Go price wasn't rendered on the pricing page when checked, so check the current plan on the tool's site, and note OpenAI says it may adjust pricing over time.
If you'd rather your practice chats not feed training, you can turn off "Improve the model for everyone" under Settings → Data controls in ChatGPT; new conversations then won't be used to train OpenAI models, and temporary chats are not used to improve models either.
Gemini (Gemini Live on mobile)
Gemini Live supports natural, free-flowing voice conversation you can interrupt, and Google lists rehearsing for important moments as a use. Two limits worth knowing: it runs in the Gemini mobile app and, per Google's help page, is not available in the Gemini web app, and you must be signed in to Gemini Apps. When you end a Live chat, the session closes and you get a transcript to review.
On plans, Google's US subscriptions page shows the Gemini app free at $0/month with a Google Account, Google AI Plus at $4.99/month, and Google AI Pro at $19.99/month with 4x higher usage access than Free, plus Gemini in Gmail, Docs and more and 5 TB storage. Prices differ by country (the Australian page shows AUD).
Is a free tier usually enough?
For most people, yes for the practice itself: a handful of spoken mock interview sessions fits inside the free tiers, since the free ChatGPT plan already includes limited voice and the Gemini app is free with a Google Account. Paid tiers mainly buy higher usage limits and longer sessions, and plans and prices can change, so check the current plan on the tool's site before you pay.
A prompt to paste
Paste this at the start of a voice session or a normal chat, and change the bracketed parts:
You are an interviewer for the role of [job title] at [company or organisation type].
Ask me one question at a time and wait for my answer before asking the next.
Question order: one opening question, then [role-relevant skill, e.g. stakeholder
management], then behavioural questions, then one question about a mistake I made.
After each of my answers:
1. Tell me whether the answer was too long, too short or about right.
2. Check my answer against STAR: Situation, Task, Action, Result — including what
I learned. The Action should be about what I personally did and take up about
60% of the answer; the Result about 10%. Tell me where my answer broke this.
3. Rewrite my answer in STAR form, in my own words, and keep it short and
conversational rather than rehearsed.
4. Ask me one follow-up question a real interviewer would ask about that answer.
Ask me [number of questions] in total. At the end, name my weakest answer and the
one thing I should fix first. Do not praise me; be specific.
Two follow-ups that work well once you have a first pass:
Now run the same interview again, but be harder: interrupt me, ask for numbers, and
challenge every vague claim I make.
Here is my CV: [paste CV or a short summary]. Ask me the three questions my CV makes
most likely, and tell me which of my examples is weakest for this role.
How far AI gets you
AI assists. It can ask plausible questions, keep the back-and-forth going, and show you where your answer stops being STAR-shaped, but it has no idea whether your example is actually strong for that employer or that interviewer, and both OpenAI and Google warn their tools can produce inaccurate output — so treat the feedback as a rehearsal partner, not as a verdict.
What you still do
- Check the facts before you say them out loud. OpenAI says ChatGPT can produce incorrect or misleading outputs, including fabricated quotes or citations, and advises verifying important information (OpenAI Help Center). Google advises double-checking responses and not relying on them for professional advice. Anything you repeat in an interview — dates, numbers, what a project delivered — is yours to confirm.
- Keep the examples yours and short. The UK National Careers Service describes STAR as Situation, Task, Action and Result, including what you learned, and advises keeping examples short, sounding conversational rather than rehearsed, and being ready for follow-up questions (STAR method). The University of Washington Career & Internship Center adds that Action should focus on what you personally did and account for about 60% of the answer, with the Result roughly 10% (UW Career & Internship Center).
- Practise with a person too. The National Careers Service recommends practising answers to common interview questions, possibly with someone you trust, and preparing your own questions to ask (Interview tips). A colleague, mentor or careers adviser will catch things a voice model won't.
- Decide what you're willing to say to a tool. Turn off "Improve the model for everyone" under Settings → Data controls in ChatGPT, or use a temporary chat, if you'd rather new conversations not be used to train OpenAI models. Remember that Live and Advanced Voice audio clips are retained for 30 days.
- Go to an official source or a qualified person for the decisions that bind. Pay and award rates, contract terms, work rights and visa questions, and health or disability adjustments at interview should come from the employer's HR, an official government source, or a licensed professional — not from an AI summary of them.
Common questions
Is the free plan enough for AI interview practice?
Usually yes for a few sessions. The free ChatGPT plan includes limited voice, and the Gemini app is free with a Google Account; paid tiers are about higher usage limits. Check the current plan on the tool's site for today's limits and prices.
Can AI tell me what this employer will actually ask?
No. It can generate plausible ChatGPT interview prep questions for a role and pick at the weak spots in your answers, but it doesn't have that employer's question set, and its feedback can be wrong.
Will the interview be realistic if I use voice?
Closer than typing, but not the same. ChatGPT's Live mode and Gemini Live both let you interrupt and keep talking, which is the part of an interview that matters most to rehearse; the missing part is a human reading your face, your pace and your hesitation.