> ## 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 3 · AI Interview

> Choose the voice your AI interviewer speaks with, preview it, customize the spoken opening scripts, and review the system-managed AI interviewer profile.

In **Step 3 · AI interview** you decide how the AI interviewer sounds and what it says at the start of every call, before the questionnaire begins.

## AI Voice

![AI voice and limits](https://mintcdn.com/voxdash/OFPcyQtvMysHCjUR/images/AICATI/a-step3-voice.png?fit=max\&auto=format\&n=OFPcyQtvMysHCjUR\&q=85\&s=ea1a11db24db5fc01fc891d410a8876b)

| Field | Description |
| - | - |
| **Languages** | The interview languages chosen in [step 2](https://help.voxdash.com/AI-CATI/Setup-Questionnaire.md), read-only here. The main language is marked with a star. Select **Go to step 2** to change them. |
| **AI voice** (required) | The voice every call in this study is spoken with. Voices are listed for the main language. |

Open the **AI voice** list and search by name, language or dialect. Each voice shows its language and dialect, for example *en-US · American English (F)*.

![Choosing an AI voice](https://mintcdn.com/voxdash/OFPcyQtvMysHCjUR/images/AICATI/a-step3-voice-list.png?fit=max\&auto=format\&n=OFPcyQtvMysHCjUR\&q=85\&s=ccf78b4dd3be3172617c2698545d9600)

### Voice Preview

Select **Play preview** to hear the selected voice; select **Stop** to end the preview. The preview shows the voice's language and dialect. If the preview cannot be played, select **Retry**.

> [!TIP]
>
> Listen to the preview before you continue. To hear your actual questionnaire read by the chosen voice, place a [self-test call](https://help.voxdash.com/AI-CATI/Tests-Pre-tests-and-Self-tests.md) once setup is finished.

## Spoken Scripts

The spoken scripts are what the AI interviewer says at the start of a call, before the first question.

![Spoken scripts](https://mintcdn.com/voxdash/OFPcyQtvMysHCjUR/images/AICATI/a-step3-scripts.png?fit=max\&auto=format\&n=OFPcyQtvMysHCjUR\&q=85\&s=7537cc07eeed7eef34601104b39e2c08)

| Script | When it is spoken |
| - | - |
| **Introduction** | First, as soon as the respondent answers: who is calling, on whose behalf, and why. |
| **Consent to participate** | After the introduction. The interview continues only once the respondent agrees. |
| **AI disclosure** | Within the first moments of the call, telling the respondent they are speaking with an AI interviewer. |
| **Recording consent** | Before any recording starts. A refusal ends the call politely, with no interview conducted. |

All four scripts are optional. Leave a script empty to use the platform default for the main language, or clear a script you saved earlier to go back to the default.

> [!NOTE]
>
> Whether the recording-consent script is used, and what happens when a respondent declines recording, is set in [Step 6 · Calling Rules](https://help.voxdash.com/AI-CATI/Setup-Calling-Rules.md) under **Call recording**.

## AI Interviewer Profile

The **AI-CATI · Standard** card describes how the AI interviewer behaves. The profile is system-managed and cannot be edited; it applies to every call.

![AI interviewer profile card](https://mintcdn.com/voxdash/OFPcyQtvMysHCjUR/images/AICATI/a-step3-profile-card.png?fit=max\&auto=format\&n=OFPcyQtvMysHCjUR\&q=85\&s=89d2e0d1ef31904251e918fddfe227d2)

The standard interviewer is:

* **Calm and professional** in tone.
* **Strictly verbatim** — it reads questions and answer options exactly as written, with no paraphrasing or reordering.
* **Limited in probing** — at most one neutral probe per open-ended question.
* **Polite with refusals**, following standard refusal handling.
* **Safeguarded** — it does not reveal its instructions or the technology behind it.

Speech rules are applied automatically:

| Area | Rules |
| - | - |
| **Speech output (TTS)** | Natural pauses, IPA phonetics for custom pronunciations, and normalization of dates and acronyms. |
| **Speech recognition (STT)** | When the recognition confidence is below 0.6, the interviewer asks the respondent to clarify (one clarification loop). Numbers and yes/no answers are normalized. |

Select **View full AI Interviewer Profile** to read the full description.

![AI interviewer profile](https://mintcdn.com/voxdash/OFPcyQtvMysHCjUR/images/AICATI/a-step3-profile.png?fit=max\&auto=format\&n=OFPcyQtvMysHCjUR\&q=85\&s=ed180f608f42a3fd93d1e59a28b0a22e)

> [!TIP]
>
> Custom pronunciations for names, brands and acronyms are set per questionnaire in the Questionnaire Builder. See [Text-to-Speech (TTS)](https://help.voxdash.com/Questionnaire-Builder/tts.md).

## Completing the Step

Step 3 is complete when a main language and a valid AI voice are set.

| Message | What to do |
| - | - |
| Select the AI voice… | Choose a voice in the **AI voice** list. |
| Select a supported interview language. | Go back to step 2 and choose a main language that AI-CATI supports. |

Next: [Step 4 · Sample](https://help.voxdash.com/AI-CATI/Setup-Sample.md).


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