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

# How the Custom Challenge Prompt Works

> The Custom Challenge Prompt is a fraud-detection step that plays an audio challenge to callers Invoca's fraud model suspects are automated (robocalls or bots), so genuine callers can pass through while suspicious calls are filtered out.

## Prerequisites

This feature applies to accounts using Invoca's Call Treatment Tree features, since the challenge prompt is configured as part of your call treatment setup.

## How the challenge works

Once a call connects to Invoca, a fraud model analyzes audio cues to determine whether the caller is a human or an automated system. If a caller is flagged as suspicious, they hear a challenge prompt — by default, "Thank you for calling. Press N to continue," where N is a number from 2 to 9, randomly generated for each call. You can customize this message, but any custom message must still include an instruction to press a key to continue; "Press 5 to continue" is the default keypress if you don't specify your own.

Only callers the fraud model flags as suspicious hear this challenge. It lasts four seconds, and non-fraudulent callers hear normal ringing during this period, so most legitimate callers won't notice any difference in their call experience.

## What happens after the challenge

A caller who presses the requested key passes the challenge and connects directly through to Invoca as normal. A caller who fails to press a key, or whom the fraud model flags as fraudulent even after a keypress, is disconnected. This call is never reported in your Invoca platform or reports, and the time a call spends in the fraud detection system before connecting (or being disconnected) is never billed.

## Testing the challenge prompt

To trigger a custom challenge prompt intentionally for testing, say a random verbal cue within the first few seconds of a test call — this may cause the fraud model to treat your test call as suspicious and play the challenge prompt, letting you verify your custom message and keypress behavior.


## Related topics

- [Custom Challenge Prompts](/en/2026-02-01/api_documentation/network_integration/custom_challenge_prompts/index.md)