Generative Search: Rules for responses

We've observed that generative summaries are returning empathetic language in their responses. This typically occurs when a user's query contains phrases that imply frustration or confusion, such as:

• "I don't see…"
• "How can I get this back?"
• "This functionality disappeared"

Generative search should provide relevant, neutral information — not to interpret or respond to a user's emotional state, especially on the search results page and the agent’s Knowledge app within the Context Panel. When a summary opens with empathetic phrasing (e.g., "Sorry for the inconvenience…"), it blurs the line between search results and conversational AI behavior.

Our recommendation: Empathetic responses should be scoped exclusively to AI Agent procedures, where conversational tone and sentiment handling are appropriate. Generative summaries should remain purely informational, returning only factual content sourced from the knowledge base — regardless of how the query is phrased.

Separating these responsibilities ensures that each feature performs its intended function without overlap.

Just continuing in on this one, today the response was:

Sorry for any inconvenience—if you share which (remaining search response).

This is a response to a search query, not a conversation. Where are they meant to share the requested information? If they type the details into the search box without the original search terms, it will result in less-relevant results, and a poor experience. 

Has there been any movement in getting this corrected Zendesk?