Letting Your Widget Suggest a Help Article Before Someone Opens a Ticket
How Chat's widget suggests Knowledge articles before a conversation even starts, and why self-service deflection is worth taking seriously.
Not every visitor who opens your chat widget actually needs to talk to a person — some just haven't found the answer that's already sitting in your knowledge base. Chat's help deflection mechanic is built for exactly that moment: as a visitor types, the widget can suggest a relevant Knowledge article before a conversation even starts.
How it works mechanically
The underlying endpoint is a public, rate-limited GET request, backed by a suggestHelpArticles lookup that matches what the visitor is typing against your published Knowledge articles, and fetchHelpArticle to pull the matched article's content. Rate-limiting matters here specifically because this is a public, unauthenticated endpoint — it's reachable by anyone on your site, not just logged-in team members, so it needs its own protection against abuse independent of your normal account-level access controls.
The same capability is exposed on the public Customer API as suggest_help, which means this isn't a widget-only trick — an integration built against Chat's API (see the guide on building against the Chat API) can trigger the same suggestion logic from outside the default widget experience.
Why this is a cross-app mechanic, not just a chat feature
This is one of the more concrete places where Chat and Knowledge are actually the same product rather than two apps that happen to share a login. The widget doesn't have its own separate help-content system — it queries your real, published Knowledge articles. Write and maintain your documentation once in Knowledge, and it does double duty: it's your public docs site, and it's the material the chat widget draws from to deflect a question before a person ever has to answer it.
Turning this on
Help deflection is opt-in at the workspace level, not a default every widget ships with. If you enable it, the quality of what gets suggested is a direct function of how good your published Knowledge content is — this mechanic surfaces existing articles, it doesn't generate new ones. A knowledge base with thin or outdated coverage will produce thin or unhelpful suggestions, which is a good forcing function for keeping your documentation actually current.
What this saves you
Self-service deflection is one of the more measurable levers in support operations — industry research on knowledge-base-driven deflection points to meaningful reductions in ticket volume when it's done well. The mechanism here is concrete enough to explain exactly why: fewer visitors need to start a conversation at all if the answer surfaces before they finish typing the question.