Dragwyb Docs Dragwyb Docs
Visit Website
Updated on Oct 24, 2026

Configure AI Assistant Tone

Set the AI chatbot's tone through its system prompt, understand temperature and response-length controls, and test example replies.

In Dragwyb Click to Chat, the assistant’s tone is primarily defined by the system prompt rather than a separate tone dropdown. You can describe the voice you want—such as friendly, formal, concise, or technical—in the prompt, then test real examples and refine the wording.

Open the prompt settings

  1. In WordPress admin, open Click to Chat → AI Assistant.
  2. Select AI Engine & Prompt.
  3. In the System Persona & Instructions card, edit Custom ChatBot System Prompt.
  4. Add clear directions about voice, sentence length, vocabulary, and how the assistant should handle uncertain answers.
  5. Save the settings and test the resulting responses in Chat Preview or on a controlled test page.

Describe the tone in concrete terms

Instead of using a single word such as “friendly,” describe how that tone should appear in the response. For example:

  • Friendly and concise: use natural language, short paragraphs, and a helpful next step without unnecessary enthusiasm.
  • Professional: use clear, respectful language, avoid slang, and explain technical terms when needed.
  • Technical: provide accurate terminology and ordered troubleshooting steps, but explain assumptions and limitations.
  • Support-focused: acknowledge the visitor’s problem, ask only essential follow-up questions, and point to the correct support route when the answer is not available.

Example prompt wording: “Use a warm, professional tone. Start with the direct answer, keep routine replies brief, use numbered steps for procedures, and clearly say when the available information is insufficient. Do not invent features, prices, or policy details.” Adapt it to your business and source content.

Understand temperature separately from tone

The Chat Accuracy / Temperature control in AI Engine & Prompt adjusts response variation. Its interface ranges from 0.0 to 2.0, with 0.7 as the default. Lower values generally encourage more predictable wording; higher values allow more variation. Temperature is not a replacement for tone instructions and may behave differently across providers and models.

The same card includes Max Tokens (Response Length), which limits the maximum response length for one message. In the inspected version, the field accepts 100 to 8,192 and defaults to 500. A shorter token limit may truncate an answer, so test it with the kinds of questions your visitors ask.

Test and refine the style

  1. Ask a routine product or documentation question and check whether the response uses the intended voice.
  2. Ask for a multi-step procedure and check that the assistant uses readable formatting.
  3. Ask an unsupported question and confirm the assistant does not confidently invent details.
  4. Compare a few responses before and after changing temperature or maximum tokens, changing one setting at a time.
  5. If the style remains inconsistent, clarify the prompt with positive instructions and examples rather than adding many conflicting rules.

For the full prompt workflow, see Configure AI Assistant Instructions. For the overall provider and response settings, see AI Assistant and Response Behavior. No live tone or model test is claimed here.