Create your first AI-moderated study

How to run an AI-moderated interview in Versive: create a study, add questions, configure the AI interviewer, and share it.

In an AI-moderated interview, the AI interviewer asks your questions and probes participants' answers with follow-ups without requiring you to attend each session. In Versive, you create the study, add questions, configure the interviewer, preview the experience, and share it with participants.

Create the study

You can start a study three ways. With the AI study assistant, describe your research goals in plain language and it creates the questions, settings, and logic. You can then ask it to add or reorder questions, rewrite wording, adjust settings, or update logic. If you're testing a product, you can also import a Figma prototype and get suggested questions based on it.

If you'd rather build question by question, use the manual builder with drag-and-drop reordering and the question library. Or start from a pre-built or organization template in the Library if one already matches your research. For a walkthrough of the AI assistant path, see Create a study with AI in 60 seconds.

Add and arrange your questions

Versive has more than 20 question types. The AI conversational question provides the adaptive moderation: you write the core question, and the AI interviewer asks follow-ups based on the participant's answer.

Mix in structured types where you want comparable data: multiple choice, rating scales, NPS, ranking, matrix questions, and more. A typical study might open with a screener, use AI conversational questions for topics that need depth, and close with structured questions for the results you want to quantify. For the full list and guidance on choosing among them, see Survey question types, and when to use each.

Configure the AI interviewer

Go to Settings → General to choose how the AI interviewer runs the session. First, pick an engine. Regular AI moderates the interview and asks follow-ups at one credit per completed interview. Realtime AI (currently in beta) runs a real-time conversational interview, including voice-to-voice, at two credits per completion. The Survey engine runs a structured survey with no AI questions and uses separate, lower-cost survey credits. It appears when your plan includes a survey-credit quota.

Next, set the conversation mode. Text and voice is the recommended default because participants can choose how to answer each question. Use text only when you explicitly do not want to collect recordings. Voice with preview lets participants review the transcription; voice only disables typing; video records the participant; and voice to voice uses the Realtime engine for a live spoken conversation. See Text, voice, or video: choosing an interview mode for guidance.

Preview before you launch

Review the study in two ways before recruiting. First, open the share link and complete the study as a participant. Use this pass to check wording, media, and logic.

Second, use the Simulate tab under Launch. Pick one or more AI personas and how many times each should run, up to ten runs per persona and up to 100 simulations in a batch. Simulated responses appear in the Results tab marked as simulations, where you can review confusing questions or broken logic and include or exclude the runs from analysis. Simulations consume credits at the same rate as participant completions, so use small batches when checking wording and logic. Prototype Task, Website Task, Tree Test, Conjoint, Media Upload, Website Link, and Calendar cannot be simulated, so test those manually.

Preview and share

Versive studies do not have a separate publishing step. A new study is live when you create it, and anyone with its URL can access it. Control exposure by distributing the URL only after previewing.

The Launch tab gives you a few ways to do that, beyond Simulate:

• Link: every study gets a public share URL that anyone can use while the study is live. You can append URL parameters (for example, a source or cohort tag) to pass metadata into responses without asking participants directly; it shows up later in results and CSV exports. Those values remain visible in browser history and logs, so keep them non-sensitive and opaque rather than using an email address, name, or token.
• Send: invite specific people by email, each with their own personal link. You get merge fields, reusable templates, per-recipient delivery tracking, and automatic suppression of bounced or unsubscribed contacts.
• Recruit: source consumers and professionals through integrated Prolific and Respondent providers when you need people outside your own network. See How to recruit research participants (4 ways) for the full comparison.
• Embed: add the study to your product as an inline widget or modal using the embed SDK or a script tag, once an organization admin has allow-listed your site's domain.

After distributing the study, you can pause it at any time. Participants are redirected to a custom pause URL if you have set one. You can archive the study when the research is done, and it pauses automatically if it reaches a configured response quota.

Responses and simulated runs appear in the Results tab, where you can include or exclude simulations from analysis and generate themes and reports.