What is a conversational survey?
A conversational survey asks questions one at a time and generates follow-up probes based on what the respondent actually said.
The result is something closer to interview depth at survey scale.
A traditional survey is fixed: every respondent sees the same questions in the same order, and an interesting open-text answer goes unexplored because nobody is there to ask "why?" A conversational survey probes on the answer, the way an interviewer would.
Why it matters
The two research modes have always traded depth for reach. Interviews go deep with 20 people; surveys reach 500 shallowly. The interesting part of most survey datasets is the open text, and that is the part with no follow-up. Probing changes the economics of that trade.
Example
A respondent writes "the reporting is frustrating." A fixed survey records five words. A conversational survey asks what they were trying to do, and gets the workflow, the workaround, and the reason: the material an interview would have produced.
In UpSight
A UpSight survey runs in one of three modes: a typed form, a chat, or a voice call, all from a link with no login. In chat and voice, an AI interviewer asks one question at a time, and you set how far it may probe. Strict asks the question as written and adds nothing; it will only re-ask if the answer is vague or off-topic. Moderate allows one brief follow-up when an answer is interesting, unclear, or too thin to use. Adaptive probes deeper whenever an answer touches your research goals. Strict is the default, so the probing is something you turn on. Every answer lands as evidence with the respondent's name on it, in the same themes as your interviews.
Related terms
FAQ
- How is a conversational survey different from a chatbot?
- A chatbot is usually there to route or support. A conversational survey is a research instrument with defined objectives and output you can analyze.
- Do conversational surveys have lower completion rates?
- They ask for more effort per question but often hold attention better than a long grid of fixed questions. Test it on your own audience rather than assuming.
- Can AI follow-ups introduce bias?
- Yes. A badly prompted follow-up leads the respondent the same way a human interviewer can. Read the generated probes for leading language.