Use response control to prevent your agent from saying things it should not — offensive language, competitor names, filler phrasing, or mispronounced terms. Without these controls, the LLM can produce unexpected outputs that undermine trust or violate compliance requirements.
Stop keywords
Block or log specific phrases in agent output (content moderation).
Pronunciations
Control how your agent pronounces specific terms (voice tuning).
Related pages
Analytics dashboard
View stop keyword metrics and trends.
Actions
Configure actions triggered by Managed Topics.