Questions,answered.
How testee works for your team and your candidates, what the AI does and doesn't do, and how billing and your data are handled.
Getting started
testee runs structured interviews with an AI interviewer. You build an interview from blocks, each with the expectations you want covered; candidates take it by voice or text whenever suits them; your team gets a report of how each expectation was covered, with the quotes behind it. How it works.
Sign up and you get 14 days free, with no card. The trial includes 150 interview minutes, interviews up to 30 minutes, 3 team members and 1 company.
Yes. The dashboard works on phones and tablets and installs as an app. Turn on push notifications to get results, mentions and decisions on the device; on an iPhone or iPad, push works from the app added to the Home Screen.
Every block type, the builder, templates and the question library, 3 interviews generated from a job description or a role title, 100 assistant messages, 60 free preview minutes, full reports, CSV exports, Slack notifications, the audit log, the public API and one webhook endpoint. ATS connections and the embed come with the paid plans. Compare the plans.
Paste a job description, or just type the role title, and review the requirements before the AI drafts the interview, with any of the seven block types and in any of the 13 interview languages. Or start from a template or a blank interview, add blocks from the library, or build it with the AI assistant, which proposes each block for you to confirm. Nothing is published until someone on your team publishes it. See the builder.
Yes. Preview runs the interview exactly as a candidate would take it, under your own name. Previews use the plan's free preview minutes first.
The interview
An AI interviewer, called Mira unless you rename it, with the voice you pick. It asks your questions, follows up on the answers and keeps to each block's time. It never reveals the expectations or correct answers, even when asked.
By voice, with live captions of both sides on screen. A candidate who can't or would rather not use a microphone picks text mode before starting and types the answers instead.
13 languages. Arabic, Dutch, English, German, Hebrew, Italian, Spanish and Urdu are tested: our interviewer and analysis evaluations have run in them and met their targets. The other 5 (Brazilian Portuguese, French, Hindi, Indonesian and Simplified Chinese) are in beta: speech recognition supports them, but the evaluations haven't met their targets in them yet. The interviewer speaks and transcribes in the interview's language, and the AI drafts interviews in any of them; the dashboard assistant replies in your interface language (Account → Interface language), whatever language you write in. The report is written in English, with the candidate's quotes kept as they were said.
As long as you build it, up to 60 minutes (30 on the trial). Each block has its own time limit, and the interviewer starts wrapping a block up when most of its time is used. Interviews generated from a job description or a role title are 15, 30, 45 or 60 minutes. For how long each kind of round should be, see structured interview best practices.
Seven block types: conversation with follow-ups, quizzes, coding (the candidate picks the language; nothing runs the code), whiteboard design, role-play with a character the AI plays, review of code or a document with planted problems, and written response. All the block types.
A browser on a computer, tablet or phone, and a microphone and speakers or headphones, or text mode instead of the microphone. A camera, and a computer to share the screen, only when your company records them.
The timer and the interviewer pause until the candidate is back, for up to 5 minutes per pause. Reloading the page continues the same interview.
If the AI service is at capacity when a candidate starts, they're asked to try again in a few minutes. Nothing is used: the interview hasn't started, no minutes count and the link stays valid.
Candidates
Add them one by one or import a CSV of up to 5,000, and invite them to a published interview; or invite them from your ATS or the API. They get an email with a link and, by default, sign in with a one-time code sent to that address, so the link alone isn't enough.
Yes. Before starting they read the interview terms, which say what is recorded and that it is analyzed by AI against your expectations, and, when you record camera or screen, the recording notice. Nothing starts until they agree, and their consent is stored with the interview.
Text mode for anyone who can't use a microphone, captions that are always on, and integrity notes that treat pastes and focus changes from assistive tools as expected. And extra time: a company can give a candidate who asks 25%, 50% or 100% more on every time limit, as long as the interview stays within 60 minutes. Accessibility.
Your company's name and branding, the interview's parts by type and length, and each question when its part begins. Never your expectations, the correct answers, a role-play's hidden brief or any score. Afterwards they can rate the interview.
Not with the same invitation: once started, it can't be restarted, though a reload or a dropped connection continues the same interview. Your team can send a new invitation if it decides to.
No. The report is for your hiring team. A candidate who asks for their data can be sent it: your team exports it as a file in one step.
Results
How each expectation was covered (covered, covered after a prompt, partial, incorrect or not covered) with the exact quotes behind it, strengths and gaps, a summary, 5 to 8 suggested follow-up questions for your next round, the transcript in sync with the recording, and a coverage score. Download it as a PDF, or export the results list as CSV.
No. Coverage is always labeled for review, not a hiring decision. People on your team set the decision (advance, hold or reject) with a note, and its history of changes is kept. Many teams run the structured rounds with the AI interviewer and keep a short final interview for themselves, with the report's suggested follow-up questions. How to read the report.
We measure it: on our test set of sample interviews labeled by people, the analysis agrees with the reviewers' verdict on about 90% of expectations. Still, like any AI, transcription and analysis can sometimes make mistakes, so every point links to its evidence (the quotes, the transcript and the recording) for a reviewer to check. The analysis judges only what was said and done against your expectations, never emotion, accent, fluency or personality. How we keep AI fair.
Comments with @mentions on each result, a team decision with its history, the PDF shared by email or the summary and its link sent in WhatsApp, Telegram, Viber or LINE to people involved in the decision, and notifications in the app, by email, in Slack or as a push on your phone when results are ready.
Integrity
With integrity signals, off by default (a company turns them on, told to the candidate before they consent; in the EU, the EEA and the UK reading browser events needs the candidate's consent, so we advise leaving them off there): the tab hidden or losing focus, copy, cut and paste (where and how many characters, never what), offline periods, reconnects and reloads, a second connection and a change of browser. The report shows them on a timeline per block, in plain words.
No. They're signals, not proof, and never feed coverage, the score or the summary. A reviewer reads them alongside the recording and can mark each one explained or dismissed.
No face analysis, eye tracking, voice biometrics or lockdown browser. A company can turn on camera and screen recording; the candidate is told and agrees first, and the video is only for people to watch: no AI processes it.
A second device, someone off screen, or apps built to stay invisible to the browser. That's why every report suggests follow-up questions to check the candidate's understanding in a live round.
Integrations
Workable, Greenhouse, Teamtailor, Recruitee and Breezy HR, on every paid plan. Link a job's stage to an interview: a candidate moved into that stage is invited once, and their results go back to the ATS as a note. Integrations.
Yes, on every plan, the free trial included: a REST API for interviews, candidates, invitations, interview status and results, with scoped API keys, plus signed webhooks when an invitation is created or opened, an interview starts or completes, and results are ready. API reference.
Yes. The embed SDK frames the interview in your page and tells it when the interview is ready, started or completed. On every paid plan, for the sites you allow.
Yes. Send notifications such as results ready or an interview completed to the Slack channels you choose, type by type.
Yes, on every plan: import candidates from a CSV of up to 5,000 rows, and export the results list as CSV (on a paid plan, your invoices too).
Privacy and security
On Google Cloud in the European Union. The voice interviewer, transcription and analysis run on OpenAI's API in the United States; OpenAI doesn't train on it and keeps it for up to 30 days to detect abuse. Where data lives.
No. We don't use your data or your candidates' data to train AI models.
Google Cloud and Firebase, Cloudflare Turnstile, OpenAI, Anam, Resend, Stripe, Web Push services (Google, Apple, Mozilla, Microsoft), Plausible Analytics, Upstash. What each one does, the data it gets and where: Sub-processors.
By default, recordings are deleted 180 days after the interview and candidates with no activity are erased after 365 days; each company can change both. Transcripts and reports stay until the candidate is erased. Privacy policy.
Erase the candidate in one step from the dashboard or the API: their invitations, interviews, recordings, reports and comments are deleted in every service. You can export their data as a file first, for a data request.
Encryption in transit and at rest, two-factor authentication for every user who turns it on, single sign-on (SAML or OIDC) on Scale and Enterprise, and an audit log of who did what, kept 365 days. Security.
Billing
A monthly plan by interview minutes: Starter includes 450 a month, Growth 1,500 and Scale 3,900; Enterprise is by contract. Minutes are shared by every company in your organization. Pricing.
The time a candidate actually spends in the interview, rounded up to the minute, and never more than the length the interview was built for. An interview that never connected costs nothing.
New interviews stop until the next month, and nothing extra is charged. An owner can turn on extra usage to keep going, billed per minute at the plan's price under a monthly cap they set, or with no cap. Interviews already running always finish.
The trial lasts 14 days or 150 interview minutes, whichever comes first. Then new interviews stop until an owner chooses a plan. You're never charged automatically: the trial doesn't take a card.
Yes. An upgrade applies at once: you pay the prorated difference for the rest of the billing period and get the new plan's full allowance, minus what you've already used. A downgrade takes effect at the end of the billing period. The Subscription page shows the price of every change before you confirm it.
An owner cancels on the Subscription page. The plan keeps working until the end of the period you've paid for, and you can resume it until then.
US dollars, euros or pounds sterling, each with its own fixed prices. The owner picks one before the first purchase; after that the organization is billed in it.
Every month's invoice is on the Subscription page's billing history, with its PDF and a CSV export. Payment methods, billing details and tax IDs are managed in Stripe's customer portal.
Prices exclude tax. Stripe sells you the monthly and yearly plans as our reseller (merchant of record): it adds VAT or sales tax at checkout from your billing address, takes your VAT ID if you buy as a business, and is the seller on your receipts and invoices. Enterprise is sold and invoiced by us under its contract.
Yes. Starter, Growth and Scale can be billed yearly for 2 months free: you pay 10 times the monthly price once a year. Interview minutes and AI allowances still renew every month (unused ones don't carry over), and extra usage is billed monthly. Switching from monthly to annual applies at once with a prorated charge; going back to monthly takes effect when the year ends.
The subscription belongs to the organization. Owners change the plan, extra usage and payment details; admins can see the plan, usage and invoices.
Still have questions?
Write to us and a person from our team will answer.Or book a demo and see an interview end to end.