# AI Variables

Generate, extract, and transform data using natural language, with no custom scripts or locators required

Part of the Endtest documentation (Advanced). Canonical URL: https://endtest.io/docs/advanced/ai-variables

---
## Overview

**AI Variables** can be generated using the **Use AI** option within the **Set Variable** action.

They allow you to generate, extract, and transform data using natural language, without writing custom scripts or complex parsing logic.

AI Variables are especially useful when the data:

- Is dynamic or context-dependent
- Requires reasoning or aggregation
- Is difficult to extract reliably using a single locator

![endtest ai variables web page](https://endtest.io/images/ai_variables.jpg)

## Scope Options

The **Scope** defines **where the AI gets its information from**. Only **one scope** can be selected per AI Variable.

## Web Tests

### 1. General Knowledge

Uses general, real-world knowledge and does **not** depend on the application under test.

**Example prompts:**

- `Generate a random fictional mobile phone number from Sweden`
- `Generate a valid IBAN for Germany`

### 2. Web Page

When **Web Page** is selected, the AI processes:

- The current **URL**
- The **page source**
- A **screenshot** of the page

**Example prompts:**

- `Calculate the total value by adding up all the prices from the "Price" column from the table`
- `Identify the primary currency used on the page`
- `Extract the value represented by the largest segment of the pie chart`

### 3. Cookies

Extracts values from browser cookies.

**Example prompts:**

- `Extract the value of the cookie that contains the user session information`
- `Get the value of the Google Analytics _ga cookie`

### 4. Variables

Allows the AI to read, combine, and transform existing variables.

**Example prompts:**

- `Add up all the numerical values from the variables that contain the prices of the products`
- `Extract the customer ID from the nested JSON stored in the $responseBody variable`
- `Calculate the number of days between the start date and end date variables`

### 5. Test Execution Logs

Allows the AI to analyze execution logs and derive values.

**Example prompts:**

- `Get the number of total Click steps executed so far`
- `Extract the latest URL that was accessed in this test execution`

## Mobile Tests

For **Mobile Tests**, web-specific scopes are replaced with **App Screen**.

### 1. App Screen

When **App Screen** is selected, the AI processes:

- The **app screen source**
- A **screenshot** of the current screen

**Example prompts:**

- `Extract the option that is visually marked as recommended`
- `Get the name of the largest city shown on the map`
- `Extract the name of the main shopping category shown on the screen`

## Strictness Levels

When generating or extracting AI Variables, you can control how strict the AI should be.

1. **Strict**: Precise extraction. Fails if the value is unclear or ambiguous.
2. **Standard**: Balanced interpretation. Recommended for most use cases.
3. **Lenient**: Tolerant of ambiguity, formatting differences, or partial matches.

## Best Practices

- Be explicit about what value you want to generate or extract.
- Use **General Knowledge** only when the value does not depend on the application.
- Prefer **Standard** strictness for most variables.
- Use **Strict** strictness for values that affect critical test logic.