Recruiter Interview Training with AI Roleplay: Practice Structured Hiring Interviews

In many organizations, the people who run interviews are not professional recruiters: team leads, project managers, branch directors. They know the job, but run few interviews, often without a shared structure.

The pitfalls are well known: improvised questions, so answers cannot be compared; a manager who talks more than the candidate; likability replacing skills analysis; one question too many about family, age or origin, asked without thinking but prohibited. A training deck describes them, but it does not help you avoid them in the moment. You need to practice, be allowed to get it wrong, then start again. That is what roleplay with a virtual counterpart offers, alongside your existing training and HR support.

In AI-Coaching, you play the recruiter and the AI plays the candidate. The exchange runs in text or voice on the platform, and the candidate reacts to what you actually say: a vague question gets a vague answer, a well-framed open question gets them talking. You configure the persona (role, personality, AI profile) from your job description and a fictional candidate background.

Candidate profiles to replay

• The shy candidate: answers in single words. You need to build trust and prompt without interrogating.
• The talkative candidate: overflows every question. You need to redirect tactfully to cover the whole structure.
• The evasive candidate: answers sideways. You need to dig with specific examples.
• The candidate who opens up: mentions family circumstances or health unprompted. You need to react without using it.

Each profile becomes a scenario. You choose the difficulty and the target role.

An AI observer follows the conversation, gives discreet real-time advice, then produces a detailed debrief. Here is a base to adapt with your HR team; a coach or a peer can then review one passage with the manager.

Everyone knows biases in theory, but they are hard to spot mid-conversation when a position needs filling. In simulation, a persona can trigger them, and you can then see them in the debrief.

Three biases come up often: similarity bias (preferring the candidate who resembles us), the halo effect (one visible quality, such as fluent speech, makes you forget the rest) and confirmation bias (asking only the questions that validate an idea already formed).

A useful exercise: replay the same candidate, very comfortable then hesitant, with equivalent skills, and compare your notes. It makes habits visible; we make no claim about how much bias it reduces.

Three articles of the French Labor Code frame the hiring interview (re-read on Légifrance on October 7, 2026).

• Article L1221-6: information requested from a candidate, in any form, may only serve to assess their ability to hold the job or their professional skills, and must have a direct and necessary link with the job offered.
• Article L1132-1: no one may be excluded from a recruitment procedure on grounds listed by law, including origin, sex, age, family situation or pregnancy, religious beliefs, physical appearance, surname, health status or disability.
• Article L1221-8: the candidate must be informed, before they are used, of the recruitment support methods and techniques applied to them. Results are confidential and the methods must be relevant.

In practice, the question is not only "am I allowed to ask?" but "what is the link with the job?". Here are reflexes to rehearse in simulation, to validate with your HR team (working suggestions, not legal advice).

The training goal is threefold: know what to ask, react calmly when a candidate raises a private subject, and know what you do not score. The grounds change over time: check the version in force and have a lawyer review your scripts.

A structured interview asks every candidate the same questions in the same order and scores answers against a grid set in advance. Simulation helps you master the method in five steps.

1. Start from the job description: isolate four to six skills the job truly requires.
2. Write the questions: one or two per skill, about situations the candidate has lived.
3. Set the grid beforehand: describe a strong, average and weak answer with observable signals.
4. Practice: play the interview with several candidate profiles, chain the questions, follow up.
5. Score afterwards, then compare: factual notes during, scores after, then a check against the observer's debrief.

Your job description, questions and grid can be fed into the mission (a technique known as RAG) so the virtual candidate and the debrief sound like your organization. Load only documents you have the right to use, with no data from real candidates.

AI-Coaching is a training tool. The candidate is fictional, played by the AI, and nobody is rejected or shortlisted at the end of a session. The debrief is about the recruiter's behavior, not about a real person.

The distinction matters. Annex III of the European regulation on artificial intelligence (Regulation (EU) 2024/1689) lists among high-risk systems those intended for the recruitment or selection of people, in particular to analyse and filter job applications and to evaluate candidates. A training simulator does not have that purpose. This is our reading, and it is up to your lawyer or data protection officer to confirm it for your case.

As a precaution, do not screen real candidates with the tool and do not load real CVs.

The other audience is the people sitting the interview: students looking for an internship or work-study placement, graduates, career changers. They play the candidate against a virtual recruiter.

The free demo includes a real scenario, Recrutement : premier contact téléphonique (a first phone contact with a recruiter), against a virtual recruiter. It opens from the try the demo page, in the scenarios for schools, with no account to create. An AI observer advises you, then a detailed debrief closes the session.

• Tell three achievements: context, personal role, outcome.
• React to an inappropriate question calmly, bringing the talk back to the job.

Schools and career centers can roll this out to their cohorts: see the page on the AI simulator for schools.

• Learning paths and playlists: structured interview, then evasive candidate, then a candidate who opens up.
• Skills map and multiple profiles: follow progress on listening, follow-ups and neutrality; HR designs and administers, managers practice.
• Secure link sharing: expiry and usage limit.
• Data: language model of your choice (Mistral, OpenAI, Azure, Gemini, deepseek or self-hosted), cloud or local, SCORM (LMS integration).

The same managers can also practice the annual review and difficult conversations. To scope a pilot with your job description and your grid, request a demo.