Help Center›Control What the AI Sees›When the AI Goes Off the Rails
Table of contents
  • Step 1: Check Your Toggles
  • Step 2: Clear Stale Instructions
  • Step 3: Try Smaller Steps
  • Step 4: Isolate the Problem
  • The AI Changed My Character's Gender / Made Things Up
  • The AI Keeps Repeating Generic Advice
  • Co-writer Memory Limits
  • Still Stuck?
    • Related articles

    When the AI Goes Off the Rails

    Last updated April 2, 2026

    Sometimes the AI doesn't do what you expect. Here's how to diagnose and fix it.

    Step 1: Check Your Toggles

    The most common cause of strange AI output is missing context. If your Outline is OFF or your Character notes are hidden, the AI is working blind.

    • Verify the right documents are toggled ON (blue)
    • Use the Prompt Preview (binoculars icon) to see exactly what the AI is receiving

    Step 2: Clear Stale Instructions

    Old prompts or instructions sitting above your cursor can confuse the AI. Remove or move them so the AI focuses on your current request.

    Step 3: Try Smaller Steps

    Break large requests into two or three smaller ones. Instead of "Rewrite this entire chapter in first person with more tension," try:

    1. "Rewrite paragraphs 1-3 in first person"
    2. "Now add more tension to the dialogue"

    Step 4: Isolate the Problem

    If it still feels off, flip ALL toggles OFF, generate one clean paragraph, then gradually toggle documents back ON one at a time. This helps you identify which document is influencing the AI's behavior.

    The AI Changed My Character's Gender / Made Things Up

    If the AI is inventing details about your story or contradicting your notes:

    1. Make sure your Story Bible or character documents are toggled ON
    2. Be specific in your prompts — reference character names and key details
    3. Keep your character notes concise and clear — the AI follows what it can see

    The AI Keeps Repeating Generic Advice

    If Co-writer keeps asking for details or giving generic responses:

    1. Click + to start a fresh chat — conversation history may be confusing it
    2. Reference specific documents by name: "Based on my Outline v2, suggest..."
    3. Toggle on the relevant documents before asking

    Co-writer Memory Limits

    Co-writer has a limited context window. In very long conversations, older messages get compressed. If it seems to have forgotten earlier context:

    1. Remind it of key details
    2. Reference specific documents by name so it re-reads them
    3. Start a new chat for a fresh memory — your documents are preserved

    Still Stuck?

    Switch to Support mode in Co-writer to reach our team. Include what you were trying to do, what happened instead, and any error messages you saw.

    Was this article helpful?