A synthetic voice: should you disclose it? What the EU rules say
Should you tell listeners they are hearing a synthetic voice? What the EU AI Act requires, since when and for whom, and why WeDispatch shows the notice by default.
The question comes up the moment you turn an article into audio with a voice that is not a journalist at a microphone: should you tell the listener? For a long time it was a matter of taste, or of an in-house policy. Since 2 August 2026 it is also a matter of law. This piece keeps the two apart, because they do not say the same thing: what the EU Artificial Intelligence Act requires, and what we chose to do, which goes a little further.
What the rules require, without stretching them
Regulation (EU) 2024/1689, the "AI Act", devotes its Article 50 to transparency. Two distinct obligations sit there, and they need separating because they fall on different people.
Article 50(2) is aimed at the provider of the system: whoever builds an AI system that generates synthetic audio, image, video or text must ensure those outputs are marked as artificially generated, in a machine-readable and detectable format. That is a technical obligation, on the file itself.
Article 50(4) is aimed at the deployer: whoever uses such a system to generate or manipulate content must, in the cases the article sets out, inform people that the content has been artificially generated or manipulated. That is a disclosure obligation, towards the public.
These Article 50 transparency obligations have applied since 2 August 2026. A publisher who turns articles into audio with a synthetic voice is, for this text, a deployer: the duty to inform, depending on the content and context, falls to them. The rest (exactly which content is covered, which exemptions apply) depends on your situation and is read in the text itself, not in a brochure: we will not go beyond what the article says.
What WeDispatch chose, and why the notice is on by default
Our choice predates the deadline and is simpler than the law: the notice "Read by a synthetic voice" is shown by default under the player, on every account. It was put in place on 23 September 2026. A publisher can remove it from their player settings; but until they do anything, it is there.
The reasoning fits in one sentence, and it is not a legal one: we have no reason to hide it. A good synthetic voice eventually stops being heard as one, and that is exactly what has changed in recent years. The day a listener works it out on their own, the notice you did not show turns into distrust: it is no longer information, it is something that was kept from them. Showing it by default is a refusal of that moment.
One detail captures the spirit of the setting. The notice can be removed by the publisher, except in one case: when a face is shown next to the voice (a reader's avatar, a photo). A portrait beside a synthetic voice, without the sentence that says so, suggests a person reading. There, the notice stays, whatever happens. It is the same demand for candour as our signature voice: a voice can be beautiful without pretending to be human.
Disclosing the voice is not certifying the text
A common confusion is worth clearing up: saying "this voice is synthetic" says nothing about who wrote the article. They are two separate questions, and we treat them separately. The notice under the player is about the voice. The provenance of the text (was it prepared by a language model, or not) is something else: that is the subject of voicing without generative AI, for newsrooms that ban any generated content. A newsroom may want a synthetic voice and a fully human text: the two notices sit side by side without contradiction.
To follow the logic of transparency to its end, knowing where the text goes during production matters as much as knowing it is read by a machine. Our page on French text to speech sets out the approach, and a dedicated piece details where the text goes: transparency about the voice would ring hollow if the text's journey stayed opaque.
In practice, for a publisher
Take away three things. First, the obligation exists and has applied since 2 August 2026: it is no longer a question of style. Second, as a deployer, informing the public falls to you, and the notice shown by default answers that directly, with no setting on your part. Third, if you remove it, do so knowingly, and know that it stays on as soon as a face accompanies the voice. Often the simplest and the most honest thing is to touch nothing: the voice is good, the article is good, and saying so takes nothing away from either.
Give your articles a voice with WeDispatch
This blog is itself voiced by WeDispatch. Curious how it sounds on your content?
Book a demo