Know who can actually do the job — before the first call.

Proofit runs AI screening interviews for IT hiring. A company creates a role, sends applicants one link, and receives a scored report for each candidate. Every score is backed by quotes from the candidate's answers, and each interview is scored by two independent AI evaluations. The first 3 candidates are free.

Try 3 candidates free · See a sample report · For candidates

How it works

  1. Create a role. Pick the track, specialization and level — for example Middle Frontend Developer. Optionally paste your job description: part of the questions will follow its real tasks.
  2. Send one link. Each applicant takes a 10-question AI interview in about 15–20 minutes, in English or Estonian. Vague answers get follow-up questions.
  3. Read the reports. Applicants appear in your dashboard sorted by score, with a full report for each.

How the scoring works

What the report contains

Roles covered

Software engineering (frontend, backend, fullstack, mobile, DevOps), project and product management, UX/UI design, QA and data analytics, at Junior, Middle and Senior levels.

Pricing

The first 3 candidates are free, no card needed. After that, €49 for 10 candidates or €149 for 50 (€2.98–4.90 per candidate), no VAT added. One credit per candidate who completes the interview; credits never expire, no subscription.

Frequently asked questions

How reliable is the score?

Each interview is scored against a written rubric for the role by two AI models from different providers. Every competency score must be backed by quotes from the answers, which are checked against the transcript. When the two evaluations disagree, the report is marked for human review. The score is a signal for a person, who makes the decision.

Can candidates use ChatGPT during the interview?

They are told not to. No tool can fully stop AI help on a second screen, so the report shows what can be measured — answer timing, attempts to manipulate the AI, and how first answers compare with follow-ups — and suggests questions to confirm the strongest candidates live.

Who makes the hiring decision?

A person. The score is advisory. Candidates know they are interviewed by AI and agree to share their result with the company before they start.

Can I add a practical task?

Yes. A company can add a 20–25 minute task to a role. Each task has problems placed on purpose; the report shows which ones the candidate found, their answer word for word and a reference solution. The overall score is 60% interview and 40% task; the price stays one credit per candidate.

Do I need to integrate anything?

No. You get a link to put in a job post, an email or your ATS message template.

Eesti keeles

Proofit on tehisintellekti eelintervjuude platvorm IT-värbamiseks. Ettevõte loob ametikoha, saadab kandidaatidele ühe lingi ja saab iga kandidaadi kohta hinnatud raporti. Iga intervjuud hindavad sõltumatult kaks tehisintellekti mudelit ja iga hinnet toetavad tsitaadid kandidaadi vastustest. Intervjuud toimuvad inglise või eesti keeles. Esimesed 3 kandidaati on tasuta, edasi 49 € 10 kandidaadi eest.

Practice the AI interview companies use — and see your real level

Free for candidates. 10 questions about your role, about 15–20 minutes, in English or Estonian. You get a detailed report: your score, the level you demonstrated, your strengths and what to work on next. Private by default.

Start a free practice interview

How practice works

  1. Pick your role and level. Developer, PM, designer, QA or analyst — then your specialization and the level you are aiming for.
  2. Answer 10 questions. Real work situations, not trivia. If an answer is vague, the AI asks a follow-up — just like a recruiter would.
  3. Get your report. A score from 0 to 100, the level you demonstrated, strengths, gaps and a short plan for what to improve. Download it as a PDF.

Got a link from a company?

Some companies send applicants a Proofit link instead of a screening call. You take the same kind of interview, and your report goes to that company only — after you agree. Proofit shows whether the company's identity is verified, and a person at the company makes the decision, not the AI.

Your data

  • Private by default. Practice results are visible only to you. Companies see them only if you turn on "Let companies find me" (anonymously) or apply through their link.
  • Share when you want. Make a report public to get a link for your CV or LinkedIn, and hide it again at any time.
  • Delete everything. Delete your account and all results in Settings.

Frequently asked questions

Is it really free?

Yes. Practice interviews are free for candidates. Companies pay for the interviews of their own applicants.

Can I retake the interview?

Yes. You can restart up to 3 times, and after a completed interview the same role opens again in 24 hours.

Can a person review my result?

Yes, you have the right to ask for a human review through the Help button.

Eesti keeles

Tasuta tehisintellekti harjutusintervjuu IT-rollidele: 10 küsimust, umbes 20 minutit, inglise või eesti keeles. Saad raporti oma taseme, tugevuste ja arenguvaldkondadega. Tulemused on vaikimisi privaatsed.

https://proofit.ee/candidates

Sample report: Middle Frontend Developer

Example with fictional data. This is what a company receives for each applicant who completes the interview.

Score: 74 / 100. Selected level: Middle · Demonstrated level: Middle. Recommended for a live interview.

Solid practical React experience: explains state management choices and trade-offs clearly and has debugged real performance problems. Testing knowledge is basic, and answers about accessibility were general.

Scoring check

Two independent evaluations agreed (75 and 73). Each interview is scored separately by two AI models from different providers against a written rubric for the role. Every competency score is backed by a quote from the candidate's answers, checked automatically against the transcript. If the two evaluations differ by more than 15 points or on the level, the report is marked for human review.

Competencies with evidence

  • Technical knowledge — 76: "saw every row re-rendering on each keystroke in the filter" (answer 1)
  • Problem solving — 78: "Whether the row height is fixed, because variable heights force measuring" (answer 2)
  • Communication — 81: "Input lag went from about 300 ms to under 50 ms." (answer 1)
  • Ownership and delivery — 68: "I memoised the row component and moved the filter state down" (answer 1)
  • Testing and quality — 58: "We have unit tests with Jest; I haven't set up integration or end-to-end tests." (answer 6)

Assessment signals (server-verified)

  • Implausibly fast answers (server-measured): none
  • Attempts to instruct the AI or set the score: none
  • First answers vs follow-ups: consistent
  • Tab switches / paste events (browser-reported): 1 / 0

Practical task: code review of a React component

Score 70. Found 5 of 8 problems placed in the task on purpose: XSS through dangerouslySetInnerHTML ("a product name with a script tag would run in the page"), a request on every keystroke, no error handling, sorting that mutates props, index keys. Missed: a race condition between requests, a duplicated effect, accessibility of clickable rows.

Suggested questions for the live call

  1. Walk me through the virtualised list you built: what library did you use and what broke first?
  2. How would you add an integration test for the checkout form you described?
  3. Show how you would check that a modal is usable with a keyboard only.

Try 3 candidates free

https://proofit.ee/sample-report

Pricing and purchase terms

Companies pay one credit per candidate who completes the interview; candidates use Proofit for free. The first 3 candidates are free (no card needed). Packs: 10 candidates for €49 (€4.90 per candidate), 50 candidates for €149 (€2.98 per candidate). One-time payment, no subscription; credits never expire. Prices are final, no VAT is added (the seller is not registered for VAT).

One credit includes a structured AI interview for the role, an optional practical task, and a report with a score, the demonstrated level, competency scores with quotes from the answers, two independent evaluations, integrity signals and questions for the live call.

What one credit replaces: a first screening call takes 15–30 minutes per candidate plus scheduling, so 20 applicants mean 5–10 hours of calls, with different questions each time. With Proofit the candidate spends 15–20 minutes, you spend none until you open the report, and every candidate gets the same structured questions built from your vacancy. No subscription, contract or per-seat fees.

Payment by card, Apple Pay, Google Pay or bank link; invoice by email. Credits are added within 1 working day after the payment arrives. Unused credits are refunded on request within 14 days of purchase; consumers may withdraw within 14 days, until credits are used. Seller: Victor Blumov, private individual, Võru vald, Estonia; notify@proofit.ee.

https://proofit.ee/pricing

AI practice interview for IT roles

Choose your track — software engineering, project and product management, UX/UI design, QA or data analytics — your specialization and level (Junior, Middle or Senior). The AI interviewer asks 10 role-specific questions and follows up on vague answers. The interview takes about 15–20 minutes, in English or Estonian, and is free for candidates.

After the interview you get a 0–100 score, the level you demonstrated, competency scores with quotes from your answers, strengths, gaps and a plan for what to improve. Results are private by default.

https://proofit.ee/interview

AI Transparency Statement

Proofit uses AI models through the Lovable AI gateway to run structured interviews for IT roles: a Google Gemini model asks the questions and follow-ups. The answers are scored independently by two models from different providers (Google Gemini and OpenAI GPT) against a written rubric for the role. Every competency score must be backed by quotes from the answers, which are checked automatically; if the two evaluations differ noticeably, the report is marked for human review.

The AI does not make hiring decisions, does not evaluate age, gender, nationality or any protected characteristic, and does not analyse voice, video, facial expressions or emotions. Candidates' answers are not used to train AI models. Anyone can ask for a human review of a result.

AI used to assess job candidates is a high-risk system under the EU AI Act (Annex III, employment); Proofit already informs candidates that they are interviewed by AI, keeps interview logs and offers human review.

https://proofit.ee/ai-disclosure

Privacy Policy

This policy explains what personal data Proofit processes, why, who receives it and what rights you have. Proofit (proofit.ee) is operated by Victor Blumov, a private individual in Estonia, who is the data controller under the GDPR. Practice results are private by default; companies see a candidate's result only after the candidate agrees. Candidates can delete their account and all results at any time.

https://proofit.ee/privacy

Terms of Service

Terms of use of Proofit (proofit.ee), operated by Victor Blumov, a private individual in Estonia. Companies receive 3 free interview credits and can buy more in packs; credits do not expire. Prices are final, no VAT is added. A person makes every hiring decision; a score alone is not a reason to reject a candidate. Governed by Estonian law.

https://proofit.ee/terms

AI-intervjuu IT-värbamisel: kuidas see töötab ja mida raport näitab

IT-ametikohale kandideerib sageli kümneid inimesi. Esimene sõelumiskõne kestab tavaliselt 15–30 minutit, nii et 20 kandidaadi peale kulub 5–10 tundi, ja hiljem on raske võrrelda inimesi, kellelt küsiti erinevaid asju. AI-eelintervjuu võtab selle esimese vooru enda peale: iga kandidaat vastab samadele küsimustele ja teie saate võrreldava raporti. Selles juhendis selgitame, kuidas see Proofitis töötab, mida raport sisaldab ja mida AI ei tee.

Mis on AI-eelintervjuu?

AI-eelintervjuu on lühike struktureeritud intervjuu, mille viib läbi tehisintellekt. Kandidaat avab lingi ja vastab kirjalikult umbes kümnele küsimusele oma rolli ja taseme kohta. Kui vastus jääb üldiseks, küsib AI täpsustava küsimuse. See ei ole valikvastustega test ega videointervjuu: kandidaat kirjutab oma sõnadega, nagu vastaks inimesele.

Eesmärk ei ole asendada vestlust inimesega, vaid esimene sõelumiskõne. Päris vestlus jääb alles, aga nende kandidaatidega, kellega tasub rääkida.

Miks just struktureeritud intervjuu?

Värbamisuuringutes on struktureeritud intervjuu üks paremaid töösoorituse ennustajaid. Sackett jt (2022) ülevaates oli struktureeritud intervjuu keskmine ennustusvõime (valiidsus) 0,42 ja tööproovil 0,33, mõlemad oluliselt kõrgemad kui töökogemuse aastatel (0,07). Struktureeritud tähendab, et kõigile kandidaatidele esitatakse samad küsimused ja vastuseid hinnatakse sama skaala järgi.

Sõelumiskõnes on struktuuri raske hoida: igaüks küsib veidi erinevalt ja märkmed tehakse mälu järgi. AI-intervjuus on struktuur sisse ehitatud.

Kuidas see Proofitis töötab

  • Loote ametikoha: valite rolli (näiteks backend-arendaja, QA või projektijuht), taseme ja keele (eesti või inglise) ning soovi korral kleebite kuulutuse teksti. Siis tuleb osa küsimusi teie kuulutuse päristöödest.
  • Saate ühe lingi ja jagate seda kandidaatidega kuulutuses või e-kirjas.
  • Kandidaat näeb enne algust, et intervjuu viib läbi AI, ja annab nõusoleku. Intervjuu kestab 15–20 minutit.
  • Soovi korral teeb kandidaat lisaks praktilise ülesande oma valdkonnast: leiab nõuetest vead ja kirjutab testjuhtumid (QA), vaatab üle koodi (arendaja), kirjutab SQL-päringu (analüütik), prioriseerib tööde nimekirja või lahendab hilineva projekti olukorra (projektijuht).
  • Kohe pärast lõpetamist on raport teie kontol. Kõik kandidaadid on samal skaalal, nii et neid saab kõrvuti võrrelda.

Mida raport näitab

  • Üldhinne 0–100 ja tase, mida vastused näitavad (Junior, Middle või Senior).
  • Hinne viie rollipõhise pädevuse kohta. Iga hinde juures on tsitaat kandidaadi vastusest. Server kontrollib, et tsitaat on vastustes päriselt olemas, ja ilma tõendita pädevus ei saa üle 40 punkti.
  • Kaks sõltumatut hinnangut eri pakkujate AI-mudelitelt (Google Gemini ja OpenAI GPT). Lõpphinne on nende keskmine. Kui hinnangud erinevad üle 15 punkti või näitavad eri taset, märgitakse raport inimese ülevaatuseks.
  • Tugevused ja lüngad lühidalt.
  • Aususe signaalid: ebausutavalt kiired vastused (mõõdab server), katsed AI-d juhendada või hinnet ette kirjutada, esimeste vastuste ja täpsustuste võrdlus ning brauseri teatatud vahelehe vahetused ja kleepimised. Need on signaalid tähelepanuks, mitte tõend.
  • Praktilise ülesande tulemus, kui see oli lisatud: mida kandidaat leidis ja mis jäi märkamata, koos tsitaatidega.
  • Küsimused, mida tasub kandidaadilt päris vestluses küsida.

Mida AI ei tee

  • AI ei otsusta, keda tööle võtta. Raport on soovitus ja otsuse teeb alati inimene.
  • AI ei analüüsi nägu, häält ega emotsioone ja kaamerat ei kasutata.
  • AI ei väida, et kandidaat petab. Kõrvalekalded on signaalid, mille kohta saab vestluses küsida.
  • Kandidaat teab alati, et intervjuu viib läbi AI, ja võib paluda, et tulemuse vaataks üle inimene.

EL tehisintellekti määrus (AI Act) liigitab värbamisel kasutatavad AI-süsteemid suure riskiga süsteemideks. Seepärast on läbipaistvus kandidaadi jaoks, inimese otsus ja intervjuude logi Proofitis algusest peale sees.

Millal see sobib ja millal mitte

  • Sobib, kui ühele IT-ametikohale tuleb palju kandideerijaid ja esimesed kõned võtavad värbaja või inseneri aega.
  • Sobib, kui tahate kandidaate võrrelda samade küsimuste põhjal ja näidata juhile, miks keegi järgmisse vooru jõudis.
  • Sobib värbamisagentuurile, kes peab tehnilisi kandidaate hindama ilma, et igasse vestlusse tuleks insener.
  • Ei asenda päris vestlust ega meeskonna tehnilist intervjuud. Kirjalikus intervjuus võib kandidaat kasutada abivahendeid, seepärast ongi raportis küsimused, millega vestluses teadmisi kontrollida.
  • Praegu ei sobi IT-väliste rollide jaoks: küsimused ja hindamisskaalad on koostatud tarkvaraarenduse, QA, projekti- ja tootejuhtimise, disaini ja andmeanalüüsi jaoks.

Korduma kippuvad küsimused

Kui palju AI-intervjuu maksab?

Esimesed 3 kandidaati on tasuta ja kaarti pole vaja. Edasi maksate ainult kandidaatide eest, kes intervjuu lõpetavad: 49 € 10 kandidaadi või 149 € 50 kandidaadi eest, ilma tellimuse ja lepinguta. Kandidaatidele on Proofit tasuta.

Mis keeles intervjuu toimub?

Eesti või inglise keeles. Keele valite ametikoha loomisel.

Mis rollidele see sobib?

Tarkvaraarendus (frontend, backend, fullstack, mobiil, DevOps), projekti- ja tootejuhtimine, UX/UI disain, QA ja andmeanalüütika, Junior-, Middle- ja Senior-tasemel.

Kas kandidaat saab AI-d petta?

Kirjalikus intervjuus ei saa abivahendite kasutamist täielikult välistada. Proofit mõõdab vastamisaega serveris, otsib katseid AI-d juhendada ja võrdleb esimesi vastuseid täpsustustega. Kahtlused näitab raport signaalidena koos küsimustega, millega vestluses teadmisi kontrollida.

Esimesed 3 kandidaati on tasuta ja ametikoha loomine võtab paar minutit. https://proofit.ee/try · https://proofit.ee/sample-report

Allikas: Sackett, P. R., Zhang, C., Berry, C. M., Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection. Journal of Applied Psychology, 107(11), 2040–2068. Sõelumiskõne kestus: Glassdoor.

https://proofit.ee/et/ai-intervjuu-it-varbamisel

For companies · For candidates · Sample report · AI disclosure · Privacy policy · Terms

Proofit is built in Estonia. Website: https://proofit.ee