AI recruiting in Switzerland: the practical 2026 guide

The legal framework: revFADP, CO 328b and the EU AI Act

Three bodies of law shape AI use in Swiss recruiting. First, the revised Federal Act on Data Protection (revFADP): it requires transparency about the processing of personal data, purpose limitation and data minimisation — and, for automated individual decisions with significant effect, informing the person concerned and offering human review on request.

Second, employment law: Art. 328b of the Swiss Code of Obligations limits data processing in the application relationship to information relevant to suitability for the specific role. Questions about pregnancy, religion or union membership are — narrow exceptions aside — impermissible. The Gender Equality Act additionally prohibits discriminatory selection practices.

Third, the EU AI Act: it applies indirectly to many Swiss companies (for instance when recruiting in the EU) and sets the benchmark Swiss practice orients itself by. AI systems in recruiting are classified as high-risk, with duties around transparency, human oversight and documentation. Emotion recognition in the workplace and hiring context has been banned since February 2025 — systems that analyse voice, facial expressions or "personality" are off the table.

What that means in practice: five rules for using AI

• Evaluate content only: AI may analyse what candidates say — not how they sound or look.
• Job-relevant criteria only: every question and criterion must be justified by the role (CO 328b).
• Transparency before the interview: candidates must know an AI conducts the conversation, and actively consent.
• Humans decide: AI recommends and justifies — a human makes the documented selection decision.
• Everything traceable: scores, justifications and process steps belong in an audit trail you can produce if challenged.

How structured AI pre-screening works in practice

The Suisse Talents flow shows the pattern: you create the role (an upload or link is enough), the system proposes job-relevant criteria and questions — checked by built-in question linting that flags impermissible questions. You weight the criteria and invite candidates via link.

Candidates complete the voice interview in the browser whenever it suits them — in German, English or French. Evaluation is criteria-based: every score is backed by verbatim quotes from the interview. You receive a justified ranking, compare candidates side by side and decide who to meet. Offers and rejections remain human decisions — documented and fair.

Data protection in practice: questions to ask any vendor

• Where is candidate data stored? (Suisse Talents: data stored in Switzerland; AI processing under SCC/EU-US safeguards — transparently documented.)
• What data feeds the AI analysis — and what is explicitly excluded (voice, video, emotions)?
• How long is data retained, and are there automatic deletion periods?
• Are candidates informed and asked for consent before the interview?
• Is there an audit trail that makes evaluations and decisions traceable?
• Is a data processing agreement (DPA) available?

What the right setup delivers

Used correctly, AI pre-screening solves three problems at once: speed (shortlists in days instead of weeks), fairness (same questions and criteria for everyone, evidenced by quotes) and accountability (documented justifications instead of gut feeling). Teams win back the hours currently locked in repetitive phone screens — and candidates get a process that takes them seriously: transparent, flexible and in their language.