### Answer When your voice agent is hallucinating despite knowledge base updates and ignoring prompt instructions, here are targeted fixes: ## Address High Agent Hallucination Rate Check your AI QA call sheet to identify the hallucination type [(1) ]() : - Fabrication (inventing facts) : Add the correct information to your knowledge base or system prompt so the agent has it instead of guessing - Contradiction (conflicting with provided info) : Simplify or clarify conflicting instructions in your system prompt - Confusion (misunderstanding user intent) : Break complex instructions into simpler steps, or use conversation flow nodes with focused prompts ## Improve Knowledge Base Recall If relevant knowledge base chunks aren't being retrieved : - Reduce the KB retrieval threshold and increase the number of chunks to reduce false negatives (missed relevant chunks) - Adjust these in your agent's Knowledge Base configuration; make small changes and monitor impact in later QA runs ## Strengthen Your Prompt Engineering Use sectional prompts to improve LLM comprehension [(2) ]() :

///## Style Guardrails
Be concise: Keep responses under 2 sentences unless explaining complex topics.
Be conversational: Use natural language, contractions, and acknowledge what the caller says.
Be empathetic: Show understanding for the caller's situation.

///## Response Guidelines
Return dates in spoken form: Say "January fifteenth" not "1/15".
Ask one question at a time: Avoid overwhelming the caller with multiple questions.
Confirm understanding: Paraphrase important information back to the caller.

For items the agent should NOT mention, be explicit in your prompt with clear boundaries . ## Consider Switching to Conversation Flow When agents show inconsistent behavior with single prompts, switch to Conversation Flow agents : - Each node focuses on one specific task - More predictable agent behavior - Easier to debug and optimize individual steps ## Additional Grounding Techniques - Integrate with verified databases to provide factual and verified information [(3) ]() - Use specialized datasets tailored to your specific industry to improve response accuracy - Implement real-time validation mechanisms that cross-reference AI outputs against verified sources [(4) ]()

​ To learn more, see the sources I considered: [[1] Address Metric Issues](https://docs.retellai.com/ai-qa/address-metric-issues) [[2] Prompt Engineering Guide](https://docs.retellai.com/build/prompt-engineering-guide) [[3] Troubleshooting AI Voice Agent Issues | Retell AI](https://www.retellai.com/blog/troubleshooting-common-issues-in-voice-agent-development) [[4] Guide to AI Hallucinations & How to Fix Them | Retell AI](https://www.retellai.com/blog/the-ultimate-guide-to-ai-hallucinations-in-voice-agents-and-how-to-mitigate-them) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.