A voiceonly (or voice-only) application is, as the name suggests, a software application (like a voice-interfaced chatbot, called also voicebot in this case) where the interface channel is only voice-based, without any graphical interface (GUI).
In last years, a voiceonly channel is a synonym for using smartspeakers, since when Amazon Echo device, terminal of the Amazon Alexa cloud-based system, was put on production in 2014; before, the traditional voice-only channel is of course the telephone and the IVR voice automation, still alive nowadays.
To be precise, popular smartspeakers by Amazon and Google are example of so called voicefirst devices (not just voiceonly), because there users interact with virtual assistants primary via the voice channel (through the smartspeaker), but the users can also interact with the same assistants using the chat interface on a mobile phone app.
Funny fact: the term voicefirst and especially the hashtag #voicefirst become popular few years ago on twitter, maybe used initially by Brain Roemmele. The term voicefirst become soon viral on voice / conversational AI community since then.
Digression: the step forward the voicefirst conversational user experience (UX) is a multimodal conversational experience where the channel is not just the speech recognition or the texting (input) or the synthetic voice play or a written prompt (in output), but the conversation is true multimodal when really multi-sensory, where by example the input is not just text or voice but also gaze detection, gesture detection, geolocation info, ambient sensors, etc.