AI Giving Wrong Answers
If your AI chatbot is providing incorrect, incomplete, or irrelevant answers, follow this systematic troubleshooting guide.
Common Causes
- Incomplete knowledge base -- the information the AI needs is not in your KB
- Outdated content -- your KB has old information that contradicts current reality
- Wrong AI mode -- the model is too simple for the complexity of the question
- Vague system prompt -- the AI lacks clear instructions on how to behave
- Ambiguous questions -- the visitor's question could be interpreted multiple ways
Step 1: Check the Knowledge Base
- Go to Dashboard -> Knowledge Base and search for the topic where wrong answers occur.
- If the topic is missing: add it. Use FAQ format for best results:
- "Q: What is your return policy? A: We accept returns within 30 days of purchase..."
- If the topic is present but vague: rewrite it with more specific, concise information.
- If the topic is outdated: update it with current information and re-index.
> Tip: After adding or updating content, test immediately by asking the chatbot the same question.
Step 2: Review the System Prompt
- Go to Dashboard -> Assistants -> Your Bot -> AI Configuration -> System Prompt.
- Add specific instructions:
- "Always refer to the knowledge base before answering."
- "If you are not sure about the answer, say: I am not sure about that. Let me connect you with a team member."
- "For pricing questions, always refer to the latest pricing page."
- "Never make up information about products or policies."
- Avoid overly long system prompts -- keep it focused and under 500 words.
Step 3: Upgrade the AI Mode
Different models have different capabilities:
| Question Complexity | Recommended Mode |
|---|---|
| Simple FAQ (hours, address, basic info) | Fast |
| Moderate support (troubleshooting, how-to) | Balanced |
| Complex/technical (multi-step, nuanced) | Smart |
If your Fast-mode bot struggles with complex questions, try Balanced or Smart.
Step 4: Review Conversation History
- Go to Dashboard -> Sessions.
- Find conversations where the AI gave wrong answers.
- Look at the AI's response and the context it used.
- Identify the gap: was the information missing, outdated, or was the AI misinterpreting the question?
Step 5: Use Flow Builder for Critical Topics
For topics where accuracy is non-negotiable (pricing, legal, medical, financial):
- Create a dedicated flow in Dashboard -> Flows.
- Use keyword or intent triggers to intercept these questions.
- Provide a fixed, pre-written response instead of relying on AI generation.
Step 6: Add Negative Instructions
Tell the AI what NOT to do:
- "Do not discuss competitor products."
- "Do not provide legal advice."
- "Do not promise delivery dates unless confirmed in the knowledge base."
- "If asked about topics outside your knowledge, say: I can only help with [your company] related questions."
Step 7: Test and Iterate
- Make a change (KB update, prompt edit, mode change).
- Test with 5-10 real questions.
- Check the answers.
- Repeat until accuracy is satisfactory.
Still Having Issues?
Contact support at [email protected] with:
- The question that produces a wrong answer
- The expected correct answer
- A screenshot of the conversation