AI and trust
How we use AI
Last updated: 20 September 2026
Emploify is an AI system for hiring. This page describes what the AI does in a simulation, what it never does and who makes the decisions. It is written for employers and for you as a candidate.
In short
- The person you talk with in the simulation is an AI. You are told this before the simulation starts.
- The AI scores what you say, from a transcript of the conversation. It never analyses your voice, your facial expressions or your emotions.
- Every score is shown with the quote it is based on, and every result names the AI model that produced it.
- The hiring decision is always made by the employer. The AI provides input.
The AI in the simulation
In a simulation you talk with an AI playing a role, such as a hotel guest or a customer in a shop. The role and the situation are written in advance, and the AI answers in real time in an AI-generated voice.
Before the simulation starts, you are told that you will talk with an AI and that your answers are scored by AI. The conversation is recorded with video and sound, and what you say is transcribed as the conversation happens.
The scoring
When the simulation is over, an AI model scores the transcript against research-based criteria for the role, such as Empathy & De-escalation or Communication Clarity.
- One criterion at a time. Each criterion is scored on its own, so a strong impression in one area does not colour the rest.
- Evidence before the score. The model first quotes what you said, then gives a score out of 5, against descriptions of weak and strong performance in the role.
- The same criteria in every language. A simulation is scored in the language it was held in: Swedish, English, Danish or Norwegian.
- Traceable. Every result names the AI model that produced it.
- Feedback for you. You get feedback on your simulation when it is done.
What the AI never does
- Reads emotions or personality. It draws no conclusions about emotions, personality, stress, honesty or “cultural fit”. Emotion recognition in the workplace is prohibited under the EU AI Act, and that part of the service has been removed, not merely switched off.
- Analyses voice, tone or face. Video and sound are not used in the scoring.
- Decides who moves on. The employer does.
- Sends anything without you. You can retake a simulation a limited number of times, and you are the one who sends your latest result to the employer.
- Trains a model on what you said. Data sent through OpenAI’s API is not used to train their models, and every data-sharing setting on our OpenAI organisation is turned off.
A person decides
Emploify gives the employer input: a score per criterion, the quotes behind each score and a summary. The employer reviews it and makes every decision in the recruitment. If a score is ever adjusted by hand, who made the change is recorded.
Intended purpose
Emploify is intended to help employers assess how candidates handle job-related situations during hiring, as input for a person who makes the decision. It is not intended to make hiring decisions, or to assess personality, emotions or anything other than the criteria for the role.
Limitations
The scoring is input, not a verdict. A simulation shows how someone handles a few situations, not everything the job requires.
Today we measure that the scoring is consistent, meaning the same answers get the same scores. We have not yet systematically compared it with how experienced recruiters score the same conversations, and we would rather say so than claim something we cannot show.
A transcript can be wrong if the audio is poor or someone is talking in the background. That is why the employer always sees the quotes each score is based on.
Models and providers
| Part | Model | Provider |
|---|---|---|
| The conversation in the simulation | gpt-realtime-2.1 | OpenAI |
| Transcription of the conversation | gpt-4o-transcribe | OpenAI |
| Scoring | gpt-5.5 | OpenAI |
A recruitment keeps its scoring model for as long as it runs, so every candidate for the same job is scored by the same model. Where data is stored, and for how long, is set out in our privacy policy.
The EU AI Act
Emploify is used to assess candidates in hiring. Under the EU AI Act, that makes it a high-risk AI system (Article 6(2) and Annex III).
- Since 2 August 2026, anyone who talks with an AI system must be told so, at the latest when the conversation starts (Article 50). That is why we tell you before the simulation.
- From 2 December 2027, the requirements for high-risk AI systems in employment apply. Before then we will have a quality management system and technical documentation in place, carry out a conformity assessment through internal control, and draw up an EU declaration of conformity with CE marking.
- Emotion recognition in the workplace is prohibited (Article 5(1)(f)). Our scoring is based only on what is said.
Questions
If you have questions about a simulation you took, or want a person to review your result, contact the employer you applied to. You can also email us at support@emploify.se. How we process personal data, and what rights you have, is set out in our privacy policy.