Should listeners be told when a radio commercial was made with artificial intelligence? In Austria, players from the media, advertising and communication industries have just come up with a joint answer. In November 2026 they published a set of shared principles on labelling AI-generated advertising content. Their main target: deepfakes, content that can lead audiences to believe a scene or a statement is genuine when it is not. For radio, the most sensitive issue is voices that have been artificially created or altered.
A shared reading of the EU AI regulation
These principles do not create a new law. They are meant to help the industry apply, in a consistent and proportionate way, the transparency obligations set out in the European regulation on artificial intelligence, the AI Act. Those involved include the Association of Austrian Private Broadcasters (VÖP), ORF-Enterprise, the Austrian association of magazines and trade media (ÖZV), the Association of Austrian Newspapers (VÖZ) and several advertising and communication organisations. The signatories stress that the document reflects a common understanding within the industry and is neither legal advice nor a legal opinion.
It is not the tool that matters, but the effect
First takeaway: using AI to produce a spot does not, on its own, trigger a labelling obligation. The decisive question is whether the result qualifies as a deepfake. That is the case, for example, when a real person is reproduced very realistically and the audience may believe that person actually said what is being broadcast. Importantly, there does not need to be any intention to deceive for this to apply.
In practical terms, for radio advertising:
- a synthetic voice that does not correspond to any existing person and does not create a false impression of authenticity generally does not need to be labelled;
- things are different with a realistic imitation of the voice of a well-known figure or any other real person, especially when it is used as a testimonial to promote a product or service.
Spoken notice or audio signal
The document also covers implementation. For audio content subject to the obligation, the information can be given through a spoken notice. Another option is mentioned: a single, recognisable sound signal, a kind of "earcon". This would, however, require coordination across the whole industry, followed by information for the public so that listeners know what it means. In all cases, the label must appear at the latest at the first contact with the message, clearly, unambiguously and directly linked to the content. For images and video, the industry recommends recognisable visual symbols, such as the AI icon proposed by the European Union.
The signatory organisations, however, reject blanket labelling of all advertising content produced with AI. In their view, too many notices would end up blurring the line between content that genuinely misleads listeners and harmless use of the technology. Each case should therefore be assessed individually, based on the content, the context in which it is broadcast and what the audience can reasonably expect.
What it means for webradios
This framework directly concerns small operators, which are often the first to use synthetic voices to quickly produce jingles, station imaging and spots for local advertisers. A few good habits follow:
- favour generic AI voices that do not imitate anyone, to limit the risk;
- never clone the voice of a presenter, an artist or a celebrity without their consent, and include a clear notice if a real voice is reproduced;
- keep a record of the tools and production choices behind each spot, so you can justify why a label is or is not used.
Although these principles were drawn up in Austria, they are based on a common European regulation. They therefore give an idea of how the industry might interpret the transparency rules elsewhere, and deserve the attention of any webradio broadcasting advertising in Europe.
Source: RADIOSZENE