Feedback

The Feedback page collects two things from your chatbot conversations: thumbs up/down ratings visitors give on bot answers, and interest signals the AI detects automatically. Use it to spot weak answers and to find visitors worth following up with.

The feedback collection tool runs on the Basic plan or higher. Free users see the page with sample data blurred and an upgrade prompt; CSV export needs a paid plan (trials don't include export).

Filtering & Search

Filter	Description
Chatbot	Filter feedback by the chatbot that collected it.
Language	Filter by the visitor's detected language.
Interest	Filter by detected interest type. The options are built from the values actually collected, so the list reflects your real data.
Search	Full-text search across feedback content, context and session ID.

What the Table Shows

Each row is one feedback entry. The columns are:

Column	Description
Feedback	The feedback text — for thumbs ratings, this is an excerpt of the bot answer that was rated.
Interest	The AI-detected signal. Thumbs ratings land here as Positive Feedback or Negative Feedback; other signals describe what the visitor showed interest in.
Context	Surrounding conversation context for the entry.
Language	The visitor's detected language.
Date	When the feedback was recorded.

Click any row to open its full detail, or use the action buttons to view the source conversation or delete the entry.

Thumbs Ratings

When the thumbs widget is enabled, visitors see thumbs up/down buttons on bot messages. A rating is saved against the specific message:

Thumbs up — the visitor found the answer helpful (shown as Positive Feedback in the Interest column)
Thumbs down — the visitor found the answer unhelpful or wrong (shown as Negative Feedback)

Filter the Interest column by Negative Feedback to jump straight to answers people disliked, then fix the underlying data source.

Interest Signals

Beyond explicit ratings, the AI reads the conversation and tags interest signals — for example purchase intent, a specific product someone asked about, or a complaint. These are extracted from the natural flow of the chat without the visitor having to click anything.

Turning Feedback Into Fixes

Filter to Negative Feedback — start with answers visitors actively disliked
Open the conversation — read why the answer fell short
Update the data source — add or correct the content behind that answer
Use positive interest signals for sales — purchase intent or product interest is a follow-up cue

Best Practices

Enable thumbs ratings — they're the fastest direct signal of answer quality
Review weekly — small, regular passes beat a once-a-quarter cleanup
Look for repeats — the same topic rated down by several visitors is a priority fix
Act on intent signals — detected purchase intent is worth a human follow-up