Zendesk’s September update: live-chat AI translation and visible sources for AI replies

Zendesk’s September 2026 release notes describe three changes support teams will feel directly.

First, AI translation — already available for email, web forms and the API — now covers conversational channels. Translations run in real time in live chat and messaging, so an agent and a customer can each write in their own language.

Second, in the web widget and on social channels, customers can now click to view the sources an AI agent used to generate a reply, instead of seeing a source list inside the response itself.

Third, in agentic messaging, if a customer’s request is still unclear after two clarifying questions, the AI agent escalates to a human rather than continuing to ask.

For support specialists, the pattern across all three is the same: the platform is formalizing where the human enters the loop — unclear requests, source verification, and quality control of machine translation. Those checkpoints are increasingly the job description. For translators, real-time machine translation moving into support channels is another signal that routine conversational translation is being automated first, while accountability for accuracy stays human.