Customers can rate each AI answer with a 👍 thumbs up or 👎 thumbs down. Those votes roll up into a helpfulness score on your dashboard — a running read on how well the bot is actually answering.
Ratings are on the AI’s answers, not on your team. Only the bot’s replies show the thumbs; a human agent’s reply isn’t rated. So the score measures the knowledge base and the AI — it’s a quality signal you can act on, not a scorecard on your agents.
How it works
- The AI answers. In the web chat widget, each AI reply carries a small 👍 / 👎.
- The customer votes (optional). One tap. It’s never required — most answers won’t be rated, and that’s fine.
- The score updates. Votes aggregate into the helpfulness percentage in the AI-bot section of your dashboard.
Turning a 👎 into a fix
A thumbs-down is the most useful thing a customer can give you: it points at a question the AI got wrong. The fix is almost always in the knowledge sources — add or clarify the article that would have answered it, and the next person who asks gets a better reply.
Read the score together with deflection. A high share-handled-by-AI is only good news if the helpfulness score is healthy too — otherwise the bot is confidently answering wrong. The dashboard shows both side by side.
Ratings make the AI improvable instead of a black box — watch the trend, and let the thumbs-down votes tell you what to write next.