Best AI for interview preparation
You have an interview soon and want the likely questions, strong answers in your own words, and a practice run. We gave nine assistants the same job description and the same made-up background, asked for eight questions and a mock interview, then answered the first question badly on purpose and watched what each one did with it.
People also look for this as job interview, interview questions, mock interview or interview practice.
Our picks, side by side
Rated for this task · 0–3
Measured
Quality
Effort
Cost
Free tier
From
Why
★ Best overall
ChatGPT
$8/mo · free tier: usableMeasuredwaited · 0 invented · 8/8 tailored
Quality
Effort
Cost
Free tierUnlimited text chat on the default model, with limits on uploads, images, voice and deep research
From$8/mo
Why
The only tool that invented nothing. All eight outlines use the candidate's real history, it says plainly not to pretend to a renewal record the candidate does not have, and when the practice answer was weak it quoted the weak phrases, rewrote the answer from the true facts, and moved to the next question. Prep you can take into the room without checking it first.
Best for the mock interview
Kimi
$19/mo · free tier: usableMeasuredwaited · 4 invented · best coaching
Quality
Effort
Cost
Free tierUnlimited basic chat, file upload and web access on the free plan
From$19/mo
Why
The toughest and most useful coach of the nine. It led with the angle that could actually win the job - the payments startup is the same domain as a billing platform - asked the renewal-quota question as bluntly as a real interviewer would, and its feedback caught the one mistake nobody else saw, volunteering a weakness in the opener. Use it for the practice rounds, but strike the four things it says you did that you did not.
Best for facing your weak spot
Claude
$20/mo · free tier: usableMeasuredwaited · 1 invented · names the gap
Quality
Effort
Cost
Free tierChat with web search and file uploads, with a usage cap per five-hour session
From$20/mo
Why
The most honest prep. It made the missing renewal record its own question - "this is the real gap, show method instead of a fake track record" - conceded the candidate had never run a QBR, and gave three beats to rebuild the weak answer rather than a script to memorise. One soft invention in a suggested line; otherwise nothing it said about the candidate was untrue.
Evidence
0 invented facts across eight outlines and a model answer; 8 of 8 outlines anchored in the pasted background; asked one question and stopped; feedback quoted three phrases from the scripted answer and rewrote it without adding anything untrue.
Not on our roster
Big Interview — A paid mock-interview trainer with recorded practice and role-specific question banks, for when you want the drills without the improvisation.
Some links here earn us a commission and carry rel="sponsored". It never affects the ranking or the ratings — how we make money.
All 9 tools we tested
Rated for this task only · 0–3| Tool | Quality | Effort | Cost | Free tier | Measured | From |
|---|---|---|---|---|---|---|
| ChatGPTFree: Unlimited text chat on the default model, with limits on uploads, images, voice and deep research | waited · 0 invented · 8/8 tailored | $8/mo | ||||
| ClaudeFree: Chat with web search and file uploads, with a usage cap per five-hour session | waited · 1 invented · names the gap | $20/mo | ||||
| GeminiFree: The default model with usage limits, plus Gemini inside Maps, Gmail and Docs on a free Google account | waited · 4 invented · 4/8 tailored | $19.99/mo | ||||
| PerplexityFree: Unlimited basic searches with a weekly cap on Pro-model searches | waited · 4 invented | $20/mo | ||||
| Microsoft CopilotFree: The Copilot app on web, Windows and mobile with usage limits | waited · 1 invented · scores you | $19.99/mo | ||||
| DeepSeekFree: The full chat app with web search and file uploads; there is no paid consumer tier | waited · 2 invented · sharpest questions | Free | ||||
| GrokFree: Chat with web and X search within generous limits, no card required | waited · 3 invented · refused once | $30/mo | ||||
| QwenFree: The full chat app; there is no paid consumer tier | waited · 3 invented | Free | ||||
| KimiFree: Unlimited basic chat, file upload and web access on the free plan | waited · 4 invented · best coaching | $19/mo |
How we tested
One brief, sent once to each of ChatGPT, Claude, Gemini, Perplexity, Microsoft Copilot, DeepSeek, Grok, Qwen and Kimi on their free tiers, with account memory switched off everywhere it exists so nothing could leak in from outside the brief. The brief carried a full job description for a Customer Success Manager role and a fictional background - five years of hotel front-office management, one year at a payments startup, no HubSpot - and asked for the eight most likely questions with answer outlines built from that background, then a mock interview that asks one question and waits.
Every tool then received the same second message, a deliberately generic sixty-word answer, and its feedback was judged on whether it engaged with what was actually said. Every outline and every model answer was read against the pasted background, and any claim about the candidate that was not in it - a job done, a metric owned, a course started - was counted as an invented fact. That count, together with whether the mock interview stopped after one question, is the measured figure in the table.
All nine stopped after one question, which is not what we expected. Where they differed was in truthfulness and coaching. Six of nine put words in the candidate's mouth - hotel QBRs never run, merchants never upsold, a HubSpot course never started - and two coached the candidate to quantify account saves that never happened. Gemini gave generic frameworks for half its questions. Three tools needed handling. Grok's free tier refused once for load, Kimi's composer carried text over from an earlier refusal, and Qwen's replies do not render fully in the page.
The test was run by, or under the direction of, the one person who runs this site, with AI assistance, and every output and fact-check was reviewed by them. The surrounding text was drafted with AI assistance from our own ratings and evidence, then checked and edited by the same person. If something here is wrong, tell us and we will fix it and say so.
Questions people ask
Will it just tell me what I want to hear?
Mostly no. Eight of nine told the candidate the practice answer was too generic to use, and four said so bluntly. The bigger risk runs the other way - most of them padded the candidate's history with things that never happened, and a couple coached quantifying results that do not exist. Read every outline against your real CV before you rehearse it.
Can it actually run a mock interview, one question at a time?
Yes, all nine did - asked one question, waited, then gave feedback on the answer before moving on. Kimi and DeepSeek asked the hardest follow-ups; Gemini and DeepSeek made the candidate retry the first question instead of moving to the second, which is a coaching choice, not a fault.
Should I turn memory off before I do this?
If your assistant has memory, yes, unless you want it to answer as it knows you rather than as the brief describes you. We ran every tool in a temporary or incognito mode. Note that ChatGPT's temporary chat still uses memory by default - you have to pick "without personalization" as well.
Which one for the night before?
ChatGPT for the outlines you will actually say, because nothing in them is made up. Kimi for a hard practice round, because it interviews like someone who wants to find the hole in your story. Claude if you already know where the hole is and want help facing it rather than hiding it.