---
title: "AI Auto-Adjust A/B Testing"
description: "How to use machine learning to find the best performing copy without lifting a finger"
url: https://helpcenter.smartlead.ai/en/articles/135-ai-auto-adjust-ab-testing
collection: "Creating Email Campaigns"
updated: 2026-10-16
language: en
---

# AI Auto-Adjust A/B Testing

How to use machine learning to find the best performing copy without lifting a finger

Use the power for auto-adjustive machine learning models (the same fabric behind Facebook popular ad platform) to find the best-performing variant for your copy.

Watch how to Use AI for AB Testing Your Best Cold Emails.

### [**Watch Here**](https://youtu.be/XewNLc4-Ij8)

Why is this important? A/B testing as a marketing principle is fundamental to find the best working messaging that pierces through the logical barrier of your potential lead right through to the emotional decision-making mechanisms that influence them to click on the "buy" button.

So, with that said - here's how you can use Smartlead's AI A/B testing system to find the best-performing variant and drive up revenue for your business through booked meetings.

As you watch the tutorial below, note that there are 3 ways you can use Smartlead's updated A/B testing system.

-   Manual Equal Distribution
    
    -   In this case, you don't need to add anything else, it will equally distribute the variants across all the leads
        

**Note:** The manual percentage variant distribution works perfectly only with multiples of 10. Any value less than 10 will not be considered.

-   AI Distribution
    
    -   In this case, you need to add the winning metric that decides which variant gets the priority. You can use any of the following to make that decision
        
        -   Open Rates
            
        -   Click Rates
            
        -   Reply Rates
            
        -   Positive Reply Rates
            
    -   Lead Sample Percentage, meaning - what percentage of the lead list you uploaded do you want to be used for experimentation in finding the variant that achieves your above-defined winning metric the fastest and most consistently.
        
-   Manual Distribution
    
    -   In this case, you need to allocate the percentage distribution on a sequence variant level, it will select the first 10 that add up to 100% in total distribution for variance allocation. For this option, you can only have a max of 10 variants when doing variant distribution as we saw from experiments that a minimum of a 10% pool is needed to make any form of an effective decision in finding the best performing A/B variant.