Skip to content

Question types & logic

The Questionnaire builder is the same for every project — it doesn't matter whether you picked Survey or AI Agent. What differs is which types you lean on: a Survey typically stays closed-ended, an AI Agent guide leans on open text with probing turned on. Both live in the same editor, question by question.

The five question types

ⓘ

Info

A display-only block — intro text, instructions, a thank-you screen. Nothing to answer, just something to read before moving on.

◉

Single

Pick one option from a list. The classic single-select question.

☑

Multi

Pick any number of options. Supports Exclusive options (like 'None of the above') that clear the rest of the selection.

✎

Open

Free text. The respondent types their own answer instead of picking from a list.

▦

Grid

A matrix of the same question asked across several rows or columns — each cell can itself be Single, Multi or Open.

Answer option rules

Any answer option on a Single, Multi or Grid question can carry one of these flags:

Terminate

Selecting this option screens the respondent out and ends the session — the standard way to build a screener.

Exclusive

On a Multi question, selecting this option clears every other selection — for options like "None of the above".

Other

Adds a free-text field alongside the option, for "please specify" answers.

Static, AI Probe & AI Answered

Every question — regardless of its type — also has a second setting that controls how the AI engages with it:

Static

Asked exactly as written. No AI involvement — the default for most Survey questions.

AI Probe

The agent can follow up live based on the response — this is what makes a question conversational, on any question type.

AI Answered

Computed from the conversation instead of asked directly — the AI infers the answer rather than putting the question to the respondent.

Skip & display logic

Each question has its own Conditions: pick an earlier question, an operator, and a value. Closed questions compare against a selected option (EQUALS / NOT EQUAL); open-ended questions can also match on text (CONTAINS). Chain conditions with AND / OR, and nest groups for more complex rules — the question only shows once its conditions are met.

Preview as you build

"View Survey" opens a live test link at any point while you're building, so you can click through the questionnaire exactly as a respondent would. For AI Probe questions specifically, the AI Sandbox lets you test how the agent actually follows up before you publish anything.

Next up