AI follow-ups only where you ask for them

You can now control exactly where the AI moderator intervenes in a session. The old behavior — fire a follow-up after every survey question — is gone. In its place, two cleaner modes.

For survey sections with AI enabled, participants answer all the questions first, then the moderator reads the full set of answers and asks one well-informed follow-up that reflects everything the participant just said. One targeted probe beats five generic ones, and the context is richer because the AI has seen the whole response set, not just the most recent line.

For behavioral sections, AI is now fully optional. Toggle it off and the section ships as a static survey — the exact research questions you configured, rendered as a plain form at the end of the task. Toggle it on and the AI chat runs as usual.

This was the most-reported UX issue from the Pro beta: the moderator felt chatty, participants felt interrupted, and completion rates suffered. The new flow reads less like a chatbot that can't stop talking and more like an interviewer who actually listens before asking the next question.

Existing studies keep working. The new defaults apply to studies created from today forward.