> ## Documentation Index
> Fetch the complete documentation index at: https://help.voxdash.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Step 2 · Questionnaire

> Bind a questionnaire version published for AI-CATI to your study, choose the interview languages, and spot-check the questions and routing the AI interviewer will execute.

In **Step 2 · Questionnaire** you choose which questionnaire the AI interviewer asks. You **bind** one questionnaire version from the Questionnaire Builder to the study, choose the interview languages, and spot-check the content.

![Step 2 · Questionnaire](https://mintcdn.com/voxdash/OFPcyQtvMysHCjUR/images/AICATI/a-step2-overview.png?fit=max\&auto=format\&n=OFPcyQtvMysHCjUR\&q=85\&s=905668baf6be76fe19369d530d79da5c)

## Before You Begin

Only questionnaire versions with the status **Published for AI-CATI** can be bound. If your questionnaire is missing from the list:

1. Open it in the [Questionnaire Builder](https://help.voxdash.com/Questionnaire-Builder/Questionnaire-List.md).
2. Publish the version, then publish it **for AI-CATI**. See [Publishing a Questionnaire for AI CATI](https://help.voxdash.com/Questionnaire-Builder/Publish-questionnaire.md#publishing-a-questionnaire-for-ai-cati).
3. Return to this step and select it.

> [!TIP]
>
> If you open a new study from **Create AI-CATI** in the Questionnaire Builder (or **Run with AI-CATI** on the dashboard), that version is bound automatically after step 1.

## Selecting a Questionnaire

When no questionnaire is bound yet, the step shows **Select a questionnaire to begin**. Select **Select questionnaire** to open the picker.

![Select a questionnaire](https://mintcdn.com/voxdash/OFPcyQtvMysHCjUR/images/AICATI/a-step2-picker.png?fit=max\&auto=format\&n=OFPcyQtvMysHCjUR\&q=85\&s=571172b13dd090cf9d51a2b6e73b7d00)

In the picker:

* **Search** by name, version or language.
* Each questionnaire shows how many versions are published for AI-CATI and its languages. Expand a questionnaire to see its versions, each with the number of questions, the estimated length of interview (LOI) and the publish date.
* Select **Select** on a version to bind it. The bound version shows **Currently bound**; other versions offer **Re-bind**.
* **Open questionnaire list** opens the Questionnaire Builder in a new tab.

> [!NOTE]
>
> A version can be bound only if at least one of its languages is supported for AI interviewing. Versions without a supported language are disabled in the picker.

Selecting a version **binds** it and captures an **execution snapshot** — a frozen copy of the questionnaire with its own version hash. Later changes in the Questionnaire Builder do not change what the study asks.

## Bound Questionnaire

After binding, the **Bound questionnaire** card shows the questionnaire name and version, the version hash, the number of questions, the estimated LOI, the questionnaire languages and when it was bound.

![Bound questionnaire card](https://mintcdn.com/voxdash/OFPcyQtvMysHCjUR/images/AICATI/a-step2-bound.png?fit=max\&auto=format\&n=OFPcyQtvMysHCjUR\&q=85\&s=2c719ff5c544d45023ff7b1156efe3f8)

* **Replace questionnaire** — bind a different questionnaire or version.
* **View in Questionnaire Builder** — open the bound version in the Questionnaire Builder.

### Bind-Time Gate

When a version is bound, VoxDash checks that the AI interviewer can execute it. When every check passes, the card reads **Published for AI-CATI · snapshot intact · ready to field**. Select **View checks** to see each check.

![Bind-time gate checks](https://mintcdn.com/voxdash/OFPcyQtvMysHCjUR/images/AICATI/a-step2-gate-checks.png?fit=max\&auto=format\&n=OFPcyQtvMysHCjUR\&q=85\&s=a14ddc28b09e12349de52857b4069a2d)

| Check | What it verifies |
| - | - |
| Screening and quota questions are asked near the start | Screeners and quota questions come before the main questionnaire. |
| Every question has at least one missing-value code, such as "Don't know" | Each answerable question can record a non-answer. |
| No question is hidden forever by the questionnaire logic | Every question can be reached. |
| Logic only uses answers the respondent has already given | Conditions don't refer to later questions. |
| Every question type can be asked by the AI interviewer | The questionnaire contains no unsupported question types. |

A check marked with a warning lists the affected questions but does not block you.

If a check fails, the card reads **Bind-time gate blocked · N failing checks**, and you cannot continue. To fix it:

**1. Open the questionnaire**

Select **Fix in Builder** to open the questionnaire in the Questionnaire Builder.

**2. Fix each failing check**

Correct the questions listed under each failing check.

**3. Publish a new version**

Publish the corrected questionnaire as a new version, and publish it for AI-CATI.

**4. Re-select it here**

Return to this step and bind the new version (or select **Re-check**).

The gate also warns you if the snapshot no longer matches the published questionnaire.

> [!NOTE]
>
> A new approval and version hash are issued per run — each pre-test and the production run get their own.

### Replacing the Questionnaire

Selecting **Replace questionnaire** first asks you to confirm:

![Replace the bound questionnaire](https://mintcdn.com/voxdash/OFPcyQtvMysHCjUR/images/AICATI/a-step2-replace-confirm.png?fit=max\&auto=format\&n=OFPcyQtvMysHCjUR\&q=85\&s=d78d388b4f082cc6c068bd33d9a7fb98)

* Binding a different version recaptures the execution snapshot, with a new version hash.
* Distribution targets defined in [step 5](https://help.voxdash.com/AI-CATI/Setup-Screening-Targets.md) refer to this questionnaire's questions. If the new questionnaire's questions differ, those targets can break and must be revisited.

Select **Choose replacement** to open the picker, or **Keep current** to cancel.

> [!WARNING]
>
> The questionnaire cannot be replaced while a prepared sub-sample run holds records. Discard the prepared runs first. See [Sub-samples](https://help.voxdash.com/AI-CATI/Sub-samples.md).

## Interview Languages

| Field | Description |
| - | - |
| **Main language** (required) | The language the interview opens in. It also determines the cost estimate. |
| **Additional interview languages** | Other questionnaire languages the interview can be conducted in. |

The languages offered come from the bound questionnaire. The AI voice and default scripts in [step 3](https://help.voxdash.com/AI-CATI/Setup-AI-Interview.md) are listed for the main language.

## Questionnaire Content Review

**Questionnaire content review** is a read-only spot-check of what the AI interviewer will read. Expand the section to see one row per question:

| Column | Description |
| - | - |
| **No.** | The question's position in the call (Q1, Q2 …). |
| **Question text (what TTS reads)** | The exact text the AI voice reads aloud. |
| **Type** | The question type, for example Single-select, Multi-select, Dropdown, Open-end. |
| **Resolved variable** | The variable name the answer is stored under. |
| **Flag** | Special handling, if any. |

![Questionnaire content review](https://mintcdn.com/voxdash/OFPcyQtvMysHCjUR/images/AICATI/a-step2-content.png?fit=max\&auto=format\&n=OFPcyQtvMysHCjUR\&q=85\&s=6b6f050b0350578bf66bc47b04d297f0)

Filter the list by question type or by flag, or search it. The flags are:

| Flag | Meaning |
| - | - |
| **Conditional** | Asked only when its display condition is met. |
| **Loop** | Repeated inside a dynamic panel. |
| **Rotating** | Part of a battery whose rows are randomized at runtime. |
| **Unsupported** | Blocked by the bind-time gate. |
| **Never asked on a call** | An element that is never spoken, for example a derived value with no formula. |

Select a question to see its details: response options and codes, missing-value coding, SPSS format, measurement level, randomization, loop context and open-end capture rules. **Derived** values show their formula — they are computed from other answers and stored with the interview, but never asked.

To change anything here, edit the questionnaire in the Questionnaire Builder, publish a new version for AI-CATI and re-bind it.

## Routing Summary

The **Routing, skip, branching and randomization** card summarizes the logic the AI interviewer executes exactly as authored:

| Tile | What it counts |
| - | - |
| **Conditional questions** | Questions with display conditions. |
| **Rotating batteries** | Batteries with randomized rows. |
| **Loop questions** | Questions inside dynamic panels. |
| **calculatedValues persisted** | Derived variables stored with each response, out of all derived variables. |

Select **View questions** or **View variables** on a tile to list the items.

## Completing the Step

Step 2 is complete when a questionnaire is bound and has a main interview language. **Continue** is blocked when:

| Message | What to do |
| - | - |
| Select a questionnaire before continuing. | Bind a questionnaire version. |
| This questionnaire has blocking issues… | Fix the failing gate checks in the Questionnaire Builder and re-bind, or choose a different questionnaire. |

Next: [Step 3 · AI Interview](https://help.voxdash.com/AI-CATI/Setup-AI-Interview.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.