RTS uses AI for programme selection
Swiss public broadcaster presents an explainable recommendation system that supports, but does not replace, schedulers.
Translated from Dutch by AI
Radio Télévision Suisse is presenting an AI system within the EBU today that helps draw up the television schedule. The system weighs audience reach, cultural objectives, public value and rights, while leaving the final decision to human schedulers.
The application was developed for the programming team at RTS, the Swiss public broadcaster. According to the EBU, the system makes recommendations for programmes and explains why a particular title or combination fits a timeslot.
That distinguishes the application from a system that optimises solely for viewing figures. Public media must also weigh cultural diversity, the public remit and the availability of rights. The relative importance of those objectives may vary by time of day or programming situation.
The tool emerged from research by Théo Maëtz, Luc Guillet and Andrea Cavallaro. Their article describes a method for multi-objective decisions and uses RTS programme data. The approach seeks to adapt recommendations to the context rather than use fixed weights for every situation.
The EBU session includes a demonstration of the application and an explanation of the transition from a research prototype to use in production. Access is limited to employees of EBU members, associated organisations and approved participants.
The sources provide no figures on improved viewing figures, programming time saved or the share of final decisions influenced by the tool. For broadcasters, the immediate lesson therefore lies chiefly in the methodology: AI can make multiple editorial objectives visible, but responsibility for programming remains with the broadcaster.
Fact check Approved
Checked against the sources by a second AI agent. That check can be wrong too.
The operational description comes from the EBU announcement and the methodological claims from the public research article. Results that are not public have not been presented as facts.
- Confirmed RTS uses an AI recommendation tool for television programming. — Confirmed in the EBU’s description of the session.
- Confirmed The tool supports human schedulers and does not take over the final decision. — Explicitly stated by the EBU.
- Confirmed The method weighs multiple objectives and uses RTS programme data. — Description in the arXiv article.
- Confirmed Théo Maëtz, Luc Guillet and Andrea Cavallaro are the authors of the research. — Listed in the public research record.
- Confirmed The EBU session is limited to EBU members, associated organisations and approved participants. — The access policy is stated on the EBU page.
Editor’s note
The EBU confirms that the tool is used in RTS’s programming practice. There are no public performance figures; their absence from the article is deliberate.Sources
- From research to production: AI-assisted TV programming at RTS — EBU Technology & Innovation
- Contextual Scalarisation Thompson Sampling for multi-objective decisions in public media — arXiv
- ICPR 2026: list of accepted papers by track — ICPR 2026