AI follow-up questions: probing without a moderator

How AI follow-up questions work in a Versive AI-moderated survey: how the interviewer probes, how to steer it, and when to cap follow-ups.

An AI Question begins with one open-ended prompt, then generates follow-ups from the participant's response. A follow-up might ask for clarification, request a concrete example, or probe a vague answer. This question type is available in AI-moderated studies; the structured no-AI Survey engine does not include it. This guide explains how follow-ups are generated, how to steer them, and how to limit their number.

How the AI interviewer decides what to probe

Once a participant answers an AI Question, the interviewer evaluates the response and may ask a follow-up based on what the participant said. It does not use a pre-written branch tree. A one-word answer might prompt a request to elaborate; an answer that names a specific problem might prompt a question about that detail.

The depth and speed of probing depend on the interviewer mode, set once for the whole study under Settings → AI Interviewer:

• Quality (the default) prioritizes more natural, nuanced follow-ups and suits depth interviews.
• Fast is quicker and more scripted, which suits short studies where speed matters more than conversational nuance.

Question rephrasing is a separate setting. By default, the AI may reword the next planned question so it follows from the conversation, for example: "You mentioned onboarding was confusing. Walk me through what happened." Rephrasing changes the wording of a scripted question. A follow-up is a new question generated during the interview. You can turn off rephrasing when exact wording matters, such as for a standardized instrument, without disabling follow-ups.

Steering the follow-ups

You can guide probing for an individual question and for the study as a whole.

The AI Question has two fields:

• Add instructions: subtitle text displayed to participants under the question. Use it for context or framing, such as "Think about the last export you ran, rather than exports in general." It is not sent to the AI interviewer.
• Interviewer notes: private guidance that steers the AI's probing without appearing on screen, such as "Focus on what specifically slowed you down, rather than whether they liked the feature overall." Versive can also suggest notes based on the question.

For the whole study, Settings → AI Interviewer holds the broader controls:

• Description (up to 900 characters) sets the interviewer's tone, role, and style, such as a warm UX researcher or a neutral market-research moderator.
• Guidelines (up to 3,000 characters) are behavioral rules that apply across the whole interview. Setting guidelines replaces Versive's defaults entirely, so include all required behavior, such as probing style, topics to avoid, and how to handle an off-topic answer.

For broader inquiry, use the Exploratory question type. Give it a research goal and a turn budget, eight turns by default, and it holds a multi-turn conversation until the goal is covered or the budget runs out. A goal such as "understand why they churned" can cover more ground than one question and its follow-ups.

Example probe sequences

Take a plain AI Question with no interviewer notes: "Walk me through the last time you tried to export a report." A typical probe sequence might run:

• Follow-up 1 (clarifying): "What were you trying to do with the report once you had it?"
• Follow-up 2 (digging into a detail): "You mentioned it did not open correctly in Excel. What happened when you tried?"
• Follow-up 3 (checking completeness): "Did you find a workaround, or did you give up on exporting that way?"

Add an interviewer note such as "Focus on friction in the export flow itself, rather than pricing or plan limits." This directs the probes toward the mechanics of exporting. Notes also help with questions that tend to get thin answers. For example, "If the participant says 'fine' or 'okay,' ask what specifically made it fine" gives the AI a concrete trigger.

When to cap follow-ups

In text and voice-response studies, every AI Question can limit follow-up questions. The cap ranges from 1 to 10 and defaults to 5 when enabled. Leave the setting off for unlimited probing. Realtime voice-to-voice studies do not expose this cap; pacing is governed by the conversation and the silence threshold.

• Set a cap: use this for short studies, high-volume panels, or questions that need a predictable length across participants.
• Use no cap: use this for depth interviews where the interviewer should be able to continue probing a relevant thread.

If your study uses a duration tracker, follow-ups get their own minimum and maximum targets. These default to half of the main question's targets. The follow-up cap controls the total number of probes. Realtime voice-to-voice studies also have a silence threshold that controls how long the AI waits after a participant stops talking before responding. The threshold affects pacing but does not change what gets asked. See Text, voice, or video: choosing an interview mode for more on interview modes.

Follow-ups and the rest of your study

AI Questions work with the rest of your study's logic. To send a participant down a different path based on an answer, see Skip logic and branching, explained.

An AI Question is one of more than 20 types in the builder. Structured types such as multiple choice and rating scale collect fixed answers without follow-ups. See Survey question types, and when to use each for the full catalog.

Test probing before launch

Run one AI Question with no instructions against yourself or a colleague. Add an interviewer note where the default follow-ups miss a relevant topic, and set a cap based on the depth and completion time you need.