What is fallback message?

A fallback message is what a chatbot says when it cannot give a grounded answer. The three usual causes are that the content does not cover the question, that the question is too unclear to retrieve against, or that a rule blocks the topic. A good fallback admits the gap in one sentence and offers a next step: a rephrase, a person, an email address or a link.

It matters because the fallback is the honest alternative to a hallucination, and because visitors meet it at the worst moment, when they already have a problem. "I don't know" with no next step loses the visitor; "I don't have that in our help content, would you like me to connect you with a person?" keeps them. The 20 examples show the wording. Fallback rate, the share of messages or conversations that end in one, is also a diagnostic: a rising rate on one topic is a page that needs writing.

Where definitions split: in rule-based chatbots the fallback is the "I didn't understand" branch triggered by an unmatched keyword, and it is the main failure mode. In generative agents the fallback is a decision, the agent judged that retrieval found nothing relevant enough, and the message often doubles as the escalation offer. Count them differently: a rule-based fallback is noise, a generative one is usually a content gap. A generative agent with no fallback at all is not a success; it is an agent that answers everything, including what it should not.