AI Follow Ups

AI Follow-Up Questions let you dig deeper automatically by generating a follow-up open text question to an existing question, based directly on each respondent’s answer. This helps you capture richer context and clearer detail without manually programming additional probes.

Key Features

• Supports Open Text questions; currently unavailable for Voice/Video. When a respondent submits their answer: The AI generates a short, neutral follow-up question in your study’s primary language. If a safe and relevant question cannot be generated quickly, the follow-up is skipped to maintain survey quality.
• Users must choose between AI Follow-Up or Skip Logic per question.
• Skips automatically if a relevant or safe question cannot be generated quickly.

Why use AI follow-up questions?

• AI follow-up questions help turn vague feedback into specific, actionable insights by probing for clarity, examples, and root causes.
• They mimic a live interviewer at scale, improving response quality without adding researcher workload.

How to Use

1. From within the study builder, navigate to any open text question and simply enable the toggle to turn on an AI Follow Up question
2. (optional) Include a prompt to help guide the AI output

• The AI Follow-Up will generate a single follow-up question focused on clarification, impact, sentiment, or depth
• The AI relies on several sources of context to generate relevant questions:
  • Agent Context (if provided): Keeps the follow-up aligned with your research objectives
  • Respondents prior question and responses: Broadens contextual understanding
  • NLP signals, such as: Intent, sentiment, topics, entities
  • Prompt inputs (if applied)
• Results: View both the AI-generated probe and the respondent’s answer in the Results tab.

Best Practice

Use 1–2 AI Follow-Up questions per survey to balance deeper insights with respondent experience.

Prompt Examples

There are endless ways to configure AI-Follow Up prompts to ensure high quality questions are asked to your respondents. A few common prompts are listed below.

Note that you can test the AI output at any time in preview mode prior to launching your study.

1. Ask the respondent to share a specific example that illustrates their previous answer.
2. If the previous answer was vague or short (for example, “it’s fine” or “good”), ask what specifically makes them feel that way.
3. If the respondent expressed frustration or a negative sentiment, ask what would have made the experience better. Do not ask them to rate the severity on a scale.
4. Ask why the respondent chose that answer, but do not ask them to compare it to competitors or other products.

CSV File Export

The AI Follow-ups feature stores follow-up prompts and responses that were generated by AI for a respondent’s original answers. The CSV format for bulk export/import of AI follow-ups

• Q1a_Question_Text
• Q1a_Response
• Q1b_AI_Followup_Question_Text
• Q1b_Response

The Q1a/b refers to the parent question and response. Q2 a/b refers to the child question and response.

API Endpoint Updates

/responses endpoint:

The only difference is the addition of parentQuestionId

      {
            "visitorId": 0,
            "surveyId": 31064,
            "questionId": 241190,
            "updatedAt": "2026-01-29T18:14:18.167Z",
            "createdAt": "2026-01-29T18:14:18.167Z",
            "responseGroupUid": "d9bb8e7f-a0c5-4844-bf2f-01c6a14993fb",
            "response": "payments was confusing",
            "questionText": "can you clarify your response?",
            "parentQuestionId": 241188,
            "externalUserId": "",
            "partnerAnonymousId": "",
            "visitorUuid": "dfe1361a-0055-40f7-ab50-0e2bda8a6f06",
            "questionType": "AIDYNAMICFOLLOWUP"
        }, 

/surveys endpoint:

The only addition is parentId, and of course the type now can also include AIDYNAMICFOLLOWUP

              {
                    "id": 242496,
                    "type": "AIDYNAMICFOLLOWUP",
                    "order": 2,
                    "options": [],
                    "parentId": 242495,
                    "questionText": "Dynamic follow-up placeholder",
                    "routingOptions": [],
                    "optionsProperties": {
                        "required": false,
                        "conceptUrl": null
                    }
                }

/themes endpoint

No change