What is an AI co-host? (And why it's not a chatbot)
The short answer
An AI co-host is an artificial participant that joins your podcast or livestream the way a human guest would. It's in the room, it hears the conversation, and when you summon it, it speaks out loud in its own voice. Your audience hears it as another voice on the show, not as a text box or a summary you read later.
"AI co-host" is a phrase that has started showing up everywhere, and it's worth being precise about what it means. A lot of things wearing the label are really chatbots with a microphone bolted on.
The chatbot vs. co-host distinction
A chatbot is something you talk to. You type a question, and it types back. Even the voice-enabled ones are a private conversation between you and a model. The output lands in your ear or on your screen, not in the show.
A co-host is something that talks with you, in front of an audience. What it says is part of the content. That sounds like a small difference, but it leads to a completely different set of requirements:
- It needs a real voice in the room. Not text-to-speech you play back by hand, but audio that arrives live, in the recording, while the conversation is still happening.
- It needs its own presence. Its own tile in the call, its own name, and ideally its own recorded track, so you can edit it on its own in post.
- It needs to know when to stay quiet. A co-host that talks constantly is unusable. The good ones speak only when summoned, then get out of the way.
- It needs to be fast. A long pause kills the energy. Around a second and a half is where it starts to feel like a real exchange.
What you'd actually use one for
Once an AI can hold its own spot on a show, the use cases stop looking like "ask the robot a question" and start looking like things you'd ask a human co-host to do:
- Fact checking on the fly. Someone drops a statistic mid-conversation, and the co-host can check it and say where the answer came from without breaking the flow.
- Filling the third chair. Solo hosts get a sparring partner, with the energy of a two-person show and no second schedule to coordinate.
- Carrying a segment. Hand it a topic and let it run for a bit while you collect your thoughts.
- Being a specific character. A dry expert, a chaotic sidekick, a recurring bit. You choose the personality.
Why "in the room" is the hard part
Most AI tools are chatbots rather than co-hosts, and the reason isn't intelligence. Modern models can hold up a conversation easily. The hard part is the plumbing: getting a synthetic voice into a live recording room as a real participant, with its own audio track, in real time, on the platform you already use.
That's the problem I built maneku to solve. It joins your Descript, Riverside or Restream room as a guest, listens to the show, and speaks when you summon it, with its own tile, its own voice and its own track in the edit.
Questions people ask
Is an AI co-host the same as a chatbot?
No. A chatbot is a private conversation between you and a model. A co-host talks with you in front of an audience, and what it says becomes part of the show.
Does an AI co-host talk all the time?
A good one doesn't. maneku stays silent until you summon it, then answers and gets out of the way. A co-host that talks constantly is unusable on a real show.
Is an AI co-host the same as an AI podcast generator?
No. A generator makes a whole episode from a prompt with no recording. A co-host joins a recording you're actually making. I wrote about the difference in detail.
Which platforms does an AI co-host work on?
maneku joins Descript Rooms, Riverside and Restream Studio. The homepage always has the current list.
Want a co-host in your next recording?
maneku is in open beta on macOS. It joins Descript, Riverside and Restream as a real participant.
Get early access