Why Is the Chatbot Speaking for Me?

This happens on every roleplay AI platform to some degree. Below are some of the factors that greatly increase the chances of the AI replying on your behalf.

The greeting includes actions by the user. The greeting is technically the AI's first message, so it copies that pattern — if the greeting has you doing things, the AI assumes it's meant to write for you.

The character definition describes who {{user}} is. That tells the AI you're a character it's responsible for writing.

Your Persona’s information contains behavioral traits. When the AI reads the entire prompt to make a reply, having behavioral descriptions in your persona, gives the AI cues on what your persona ought to do, or say. The AI sees those traits as active material to fulfill, prompting it to write your character's reactions and dialogue for you.

Your replies are mostly short and/or passive. The AI is always prompted to move the story forward, but it can only do that with momentum. If it doesn’t have anything to work with, it may create momentum by taking over the user's actions to push the scene forward.

Fixes

Rewrite the greeting or reply from your character, into the chatbot character's perspective. Instead of:

{{user}} walks into the room and sits down.

write:

{{char}} watches {{user}} enter and take a seat.

Edit it out the first time it happens. Delete the part where the AI writes for you. Doing this consistently for a few messages often trains it out within the conversation.

Use a Director Mode command:

/cmd {{char}} will never write actions or dialogue on {{user}}'s behalf.

Lower the Temperature slightly in Generation Settings (I'm All In) — see article 4.11.

Add 2–3 example dialogues in the character's advanced settings, in the format you want.

Clone the chat back to before the behaviour started.

Rate the replies 1 star and select “Respond as User. It won’t affect the chat, but it provides the AI team with data that helps improve how we prompt the models."