Text Chat Survey Questions

Text Chat is a stable question type for conversational survey interactions. It can support qualitative probes, guided reflection, tutoring-style checks, interview-like prompts, or AI-assisted tasks inside a social science research survey. It requires a Standard plan or above.

• System prompt control defines how the assistant should interact with participants.
• Turn limits keep conversations comparable across respondents.
• Conversation ends chooses whether every respondent runs for exactly the same number of messages ("Fixed number of messages", the default) or the model can end the interview itself once it judges the topic is covered ("Model decides when to end"). Max turns still applies as a hard cap either way, and min turns still applies as a floor the model cannot end before.
• Endpoint configuration lets research teams connect a compatible model provider, either directly or through a saved credential.
• Message length limits set a minimum and/or maximum length (in characters, words, or sentences) for each participant message.
• Paste controls can disable pasting into the message box, and/or record blocked or allowed paste attempts in response metadata — the same paste-tracking behavior available on Short Answer.
• Track typing timing (opt-in) records keystroke count, typing duration, and longest pause across the respondent's own messages, the same typing-timing tracking available on Short Answer. It can raise the "No Typing Detected" response-quality flag.
• Transcript validation can require the finished transcript to contain (or not contain) specific words or phrases, or run a custom Python check against the full transcript text — the same validation options available on Short Answer.
• Advanced request body (in the question's advanced settings) lets you merge additional JSON fields into the request sent to the model endpoint, for providers that need parameters beyond the standard chat-completions shape.
• Transcript export supports qualitative coding and audit review, and includes which model produced each assistant reply so a later review can tell if the model changed mid-study.
• Response delay can make the interaction feel paced rather than instantaneous.

1. Step 1: Open Block Builder. Open the survey and select the block where the chat task belongs.

   Step 1: Open Block Builder

2. Step 2: Add a Text Chat question. Choose Text Chat from the question type selector.

   Step 2: Choose Text Chat question type

3. Step 3: Write both prompts. Add the respondent-facing prompt and the system prompt that controls chat behavior.

   Step 3: Write both prompts

4. Step 4: Configure the model connection. Set the endpoint, saved credential, model name, minimum and maximum turns, whether the conversation always runs the same number of messages or the model decides when to end it, message length requirements, and response delay.

   Step 4: Configure the model connection

5. Step 5: Preview with realistic responses. Use Preview to test the interaction before publishing, including awkward, short, and off-topic participant answers.

   Step 5: Preview with realistic responses

6. Step 6: Export and review transcripts. Submit a test response, confirm the transcript is suitable for coding and audit review, and check privacy expectations before live collection.

   Step 6: Export and review transcripts

Text Chat exports the chat transcript, including participant and assistant messages. Treat transcripts as qualitative research data and review privacy, consent, and institutional review requirements before collecting live responses.

• Text Response Survey Questions
• Audio Response Survey Questions for spoken instead of typed answers
• Audio Conversation Survey Questions for a spoken AI-moderated conversation instead of a typed one
• Content Blocks in Research Surveys
• Create Experimental Conditions
• Publish and Share a Research Survey