Title: AB Split Test Lite
Author: Tom
Published: <strong>November 17, 2026</strong>
Last modified: November 17, 2026

---

Search plugins

![](https://ps.w.org/ab-split-test-lite/assets/banner-772x250.png?rev=3808344)

![](https://ps.w.org/ab-split-test-lite/assets/icon-256x256.gif?rev=3808344)

# AB Split Test Lite

 By [Tom](https://profiles.wordpress.org/tomsitespot/)

[Download](https://downloads.wordpress.org/plugin/ab-split-test-lite.1.0.0.zip)

 * [Details](https://wordpress.org/plugins/ab-split-test-lite/#description)
 * [Reviews](https://wordpress.org/plugins/ab-split-test-lite/#reviews)
 *  [Installation](https://wordpress.org/plugins/ab-split-test-lite/#installation)
 * [Development](https://wordpress.org/plugins/ab-split-test-lite/#developers)

 [Support](https://wordpress.org/support/plugin/ab-split-test-lite/)

## Description

**Most website changes don’t lift your conversion rate. Many do nothing, and some
quietly make it worse. AB Split Test Lite lets you, or your AI agent, prove each
change on real visitors before you keep it.**

A new headline, a redesigned hero or a rewritten button looks like an improvement
on the day you ship it. Without a test you never find out whether it won, lost, 
or changed nothing. AI agents now write and ship site changes in minutes, so more
of those untested changes go live every week.

AB Split Test Lite closes that gap. It is a **self-hosted** A/B testing plugin with
a modern **MCP API**, built on the WordPress **Abilities API**, so Claude Code, 
Cursor, Windsurf or any MCP client can create, launch, read and end tests on your
site. Tests, visits, conversions and heatmaps are all stored in your own database,
with no external account and no traffic meter.

#### The validation loop

Give your agent a loop to work in instead of a list of edits to ship blind:

 1. **Propose.** The agent picks one change worth testing: a headline, a call to action,
    an image, a section, or a whole page.
 2. **Create the test.** One `create-test` call sets up the original, the variation
    and the goal (a page visit or an element click). Invalid tests are rejected with
    clear errors, such as a missing scope, so the agent fixes the request instead of
    guessing.
 3. **Run it.** Visitors are split 50/50 between the original and the variation, on
    your own server.
 4. **Read the result.** `get-test-results` returns visits, conversions, conversion
    rates and Bayesian confidence, so the agent knows whether there is a real winner
    or just noise. The site admin also gets an email when a winner is found.
 5. **Decide.** Keep the variation only if it won. If it lost or made no difference,
    the original stays and nothing is lost but a guess. End the test with `update-test-
    status` and start the next one.

You stay in charge: the plugin reports a winner but never changes your site on its
own.

#### Features

 * **MCP API for AI agents.** Six tools, over MCP and the matching REST API: create
   a test, list tests, read a test’s configuration, read its results, start, pause
   or complete it, and change its settings.
 * **Unlimited A/B tests.** Run as many tests at once as you need. Each test compares
   your original with one variation, and traffic is split evenly between the two.
 * **Visitor segmentation.** Choose which visitors each test includes by user role,
   device size, URL query string and share of traffic.
 * **Admin bar helper.** On any page, see the tests running there, preview each 
   version with the tested element highlighted, and open that page’s heatmaps.
 * **Trustworthy results.** Bayesian statistics, with a minimum number of visits
   and conversions before a winner is called.
 * **Self-hosted.** Tests, tracking and reporting all run on your server.
 * **No traffic meter.** No “tested page view” limit and no per-visitor billing.
 * **Heatmaps on every page.** Click maps and scroll maps, stored on your own server,
   so you can see why a variation won. You choose how many days of data to keep.
 * **Page-builder integrations.** Elementor, Bricks, Beaver Builder, Oxygen, Breakdance,
   WP Bakery and Gutenberg all get native test controls inside the builder.
 * **Privacy-first.** No device fingerprinting and no third-party services. Visitors
   are assigned to the original or the variation with first-party cookies on your
   own domain, and you can wait for cookie consent before anything is stored.
 * **Cache-friendly.** Test scripts are marked for exclusion in WP Rocket, NitroPack,
   LiteSpeed Cache, SiteGround Optimizer and other optimization plugins, so tests
   don’t flicker or break.

#### Three ways to build a test

**Point-and-click (Magic Bar).** Open any page, click the headline, button, image
or section you want to test, type your variation, and go. No selectors to hunt for
and no code to write.

**Full-page (split URL).** Send traffic between completely different pages and measure
which one converts.

**On-page.** Tag one section or block as the Original and another as the Variation,
inside your page builder or the block editor.

#### What you can test

 * Headlines, sub-headlines and body copy
 * Buttons and calls to action
 * Images and hero sections
 * Pricing and product copy
 * Entire page layouts
 * Design and style changes
 * Anything you can click on with the Magic Bar

#### Page-builder integrations

AB Split Test adds **native controls inside each builder**, so you build your variation
in the interface you already design in:

 * **Elementor**: tag an element as the Original or the Variation in the Elementor
   editor.
 * **Bricks**: tag an element as the Original or the Variation in the element settings.
 * **Breakdance**: a “Split Test” settings section on every element.
 * **Beaver Builder**: split-test settings in row and module settings.
 * **Oxygen**: split-test options in component options.
 * **WP Bakery**: tag a row or section as the Original or the Variation.
 * **Gutenberg**: split-test attributes on every block, managed from the block editor.

Each builder can also create a new test without leaving the editor.

#### Conversion goals

Pick what counts as a win for each test:

 * **Page visit**: a visitor in the test reaches the page you choose, for example
   your thank-you, order-received or sign-up confirmation page.
 * **Element click**: a visitor clicks an element you choose with a CSS selector,
   such as a buy, sign-up or call button.

#### Heatmaps and scroll maps

See _why_ a variation wins. Click heatmaps show where visitors click, and scroll
maps show how far down each page they read. Everything is recorded on every page
of your site and stored on your own server. Data is kept for 3 days by default; 
change it under **AB Split Test Lite  Settings  Heatmaps**.

#### Trustworthy results

Results use Bayesian statistical analysis, so you know when a winner is really a
winner and not just noise. A winner needs enough visits and conversions before it
is called. Filter results by device to see how the original and the variation perform.

When a running test finds a winner, the site admin gets an email about it. The test
keeps running and nothing on your site changes until you end it.

#### The A/B testing plugin your AI agent can drive

AB Split Test registers WordPress Abilities. Install the official WordPress MCP 
Adapter and they appear as **MCP tools**. The same actions are available over the**
REST API**.

 * `create-test`: create an A/B test
 * `list-tests`: list tests and their status
 * `get-test-details`: read a test’s full configuration
 * `get-test-results`: pull visits, conversions and statistical confidence
 * `update-test-status`: start, pause or complete a test
 * `update-test-settings`: change the goal, targeting and the variation

Each tool describes its inputs in full, so an agent can build a valid test without
reading documentation first. To find reliable CSS selectors, an agent can open any
page with `?abhash=1` added to the URL, which lists every visible text element with
a unique selector.

A prompt as short as _“Test a clearer headline on the pricing page, with a click
on the buy button as the goal, and tell me when there’s a result”_ is enough for
an agent to run the whole loop.

#### AB Split Test Pro

AB Split Test Pro is a separate plugin, available from [absplittest.com](https://absplittest.com/repo-up/?utm_source=wporg-lite&utm_medium=readme&utm_campaign=pro),
with more ways to measure a win: form submission goals, WooCommerce, Easy Digital
Downloads and FluentCart purchase and revenue tracking, sub-goals for multi-step
funnels, audiences and location targeting, automatic winner selection, AI test ideas,
webhooks, email reports and WP-CLI commands. Pro also runs multivariate (A/B/n) 
tests with several variations. Your tests and results carry over if you switch.

#### Third-party libraries

This plugin bundles the following open-source libraries:

 * Chart.js 4.6.1, MIT License, https://github.com/chartjs/Chart.js
 * Select2 4.1.0, MIT License, https://github.com/select2/select2
 * Driver.js 1.11.0, MIT License, https://github.com/nilbuild/driver.js
 * Tabulator 6.7.1, MIT License, https://github.com/olifolkerd/tabulator
 * heatmap.js 2.0.5, MIT License, https://github.com/pa7/heatmap.js
 * Awesomplete 1.1.7, MIT License, https://github.com/LeaVerou/awesomplete

The readable source of every minified file ships alongside it: js/chart.src.js, 
js/select2.src.js, js/tabulator.src.js, js/creator.src.js (Awesomplete) and js/bt_conversion.
js (the plugin’s tracking script, minified as js/bt_conversion-min.js). heatmap.
js ships unminified as js/heatmap.src.js. Driver.js is published only as a minified
build; its source for the bundled version is at https://github.com/nilbuild/driver.
js/tree/1.11.0/packages/driver/src.

#### Help translate

AB Split Test is translation-ready. Contribute a translation for your language at
[translate.wordpress.org](https://translate.wordpress.org/projects/wp-plugins/ab-split-test-lite/).

## Screenshots

[⌊The point-and-click Magic Bar: click any element on your page, type a variation,
and choose the goal, such as a click on a button you pick on the page.⌉⌊The point-
and-click Magic Bar: click any element on your page, type a variation, and choose
the goal, such as a click on a button you pick on the page.⌉[

The point-and-click Magic Bar: click any element on your page, type a variation,
and choose the goal, such as a click on a button you pick on the page.

[⌊Test results with uplift, confidence and conversion rates, calculated on your 
own server.⌉⌊Test results with uplift, confidence and conversion rates, calculated
on your own server.⌉[

Test results with uplift, confidence and conversion rates, calculated on your own
server.

[⌊A completed test: the winning variation, its uplift and confidence, and the projected
yearly impact.⌉⌊A completed test: the winning variation, its uplift and confidence,
and the projected yearly impact.⌉[

A completed test: the winning variation, its uplift and confidence, and the projected
yearly impact.

[⌊A test still collecting data: the current leader, its confidence so far, and an
estimate of the days until a winner can be called.⌉⌊A test still collecting data:
the current leader, its confidence so far, and an estimate of the days until a winner
can be called.⌉[

A test still collecting data: the current leader, its confidence so far, and an 
estimate of the days until a winner can be called.

[⌊Click heatmaps built from your own visitors, recorded on every page.⌉⌊Click heatmaps
built from your own visitors, recorded on every page.⌉[

Click heatmaps built from your own visitors, recorded on every page.

[⌊Heatmaps for each screen size, here on mobile.⌉⌊Heatmaps for each screen size,
here on mobile.⌉[

Heatmaps for each screen size, here on mobile.

[⌊Scroll maps show how far down each page your visitors actually read.⌉⌊Scroll maps
show how far down each page your visitors actually read.⌉[

Scroll maps show how far down each page your visitors actually read.

[⌊The admin bar helper on any page: open its heatmaps, start a new Magic Test, and
see every test running there.⌉⌊The admin bar helper on any page: open its heatmaps,
start a new Magic Test, and see every test running there.⌉[

The admin bar helper on any page: open its heatmaps, start a new Magic Test, and
see every test running there.

[⌊Click a version in the admin bar to preview it on the page, with the tested element
highlighted.⌉⌊Click a version in the admin bar to preview it on the page, with the
tested element highlighted.⌉[

Click a version in the admin bar to preview it on the page, with the tested element
highlighted.

[⌊Visitor segmentation: choose which user roles, device sizes, URL query strings
and share of traffic each test includes.⌉⌊Visitor segmentation: choose which user
roles, device sizes, URL query strings and share of traffic each test includes.⌉[

Visitor segmentation: choose which user roles, device sizes, URL query strings and
share of traffic each test includes.

[⌊Elementor: tag an element as the Original or the Variation from its Advanced tab.⌉⌊
Elementor: tag an element as the Original or the Variation from its Advanced tab
.⌉[

Elementor: tag an element as the Original or the Variation from its Advanced tab.

[⌊The block editor: an AB Split Test panel on every block, under Advanced.⌉⌊The 
block editor: an AB Split Test panel on every block, under Advanced.⌉[

The block editor: an AB Split Test panel on every block, under Advanced.

[⌊Bricks: AB Split Test controls on every element, under Style.⌉⌊Bricks: AB Split
Test controls on every element, under Style.⌉[

Bricks: AB Split Test controls on every element, under Style.

[⌊Beaver Builder: split-test settings in each module's Advanced tab. Oxygen, Breakdance
and WP Bakery have native controls too.⌉⌊Beaver Builder: split-test settings in 
each module's Advanced tab. Oxygen, Breakdance and WP Bakery have native controls
too.⌉[

Beaver Builder: split-test settings in each module’s Advanced tab. Oxygen, Breakdance
and WP Bakery have native controls too.

[⌊Every test at a glance: visits, conversions and conversion rate for the original
and the variation, with its status.⌉⌊Every test at a glance: visits, conversions
and conversion rate for the original and the variation, with its status.⌉[

Every test at a glance: visits, conversions and conversion rate for the original
and the variation, with its status.

[⌊The Developer tab: the MCP tools and REST API endpoints your AI agent can use.⌉⌊
The Developer tab: the MCP tools and REST API endpoints your AI agent can use.⌉[

The Developer tab: the MCP tools and REST API endpoints your AI agent can use.

[⌊Creating a test: choose point-and-click, full-page or on-page elements.⌉⌊Creating
a test: choose point-and-click, full-page or on-page elements.⌉[

Creating a test: choose point-and-click, full-page or on-page elements.

[⌊Heatmaps for a full-page split test: see where visitors click on each version.⌉⌊
Heatmaps for a full-page split test: see where visitors click on each version.⌉[

Heatmaps for a full-page split test: see where visitors click on each version.

## Installation

 1. In your WordPress admin, go to **Plugins  Add New Plugin**.
 2. Search for **AB Split Test Lite**.
 3. Click **Install Now**, then **Activate**.
 4. Open **AB Split Test Lite  New Test** in the admin menu and create your first test.
 5. To test an element on a page, open the page on the front end and choose **New Magic
    Test** from the **AB Split Test Lite** menu in the admin bar.

**Enable AI-agent (MCP) control (optional):**

 1. Install and activate the official **WordPress MCP Adapter** plugin. The plugin’s
    settings **Developer** tab walks you through it.
 2. Create an application password for your user under **Users  Profile**.
 3. Copy the ready-made client configuration from the Developer tab into Claude, Cursor,
    Windsurf or another MCP client.
 4. The `absplittest/*` tools are now available. The same actions are also on the REST
    API under `/wp-json/bt-bb-ab/v1/`.

## FAQ

### Is AB Split Test a self-hosted A/B testing plugin?

Yes. The testing engine, visitor tracking, heatmaps and reporting all run inside
your WordPress install and store data on your own server. There is no external account
and no SaaS backend. Heatmap data is kept in daily text files in your uploads directory.
Their names include a secret code from your site’s keys, so they cannot be found
by guessing, and an .htaccess rule blocks direct downloads on Apache and LiteSpeed.

### How many tests can I run?

As many as you like, on as much traffic as your site serves. Each test is a classic
A/B test: it compares your original with one variation, and visitors are split evenly
between the two.

### Why test a change instead of just making it?

Because most changes don’t do what you expect. A new headline or layout that looks
better often makes no measurable difference, and some lower your conversion rate
without anyone noticing. A test shows you which it is, on your own visitors, before
the change becomes permanent. That matters even more when an AI agent can make dozens
of changes a week.

### Can an AI agent like Claude run A/B tests on WordPress?

AB Split Test registers WordPress Abilities, which the official WordPress MCP Adapter
exposes as MCP tools. The same actions are on the REST API. An agent authenticated
as a user with the right permissions can create tests, list them, read results, 
change their settings, and start, pause or complete tests.

### How do I A/B test an Elementor, Bricks or Beaver Builder page?

AB Split Test adds native controls to Elementor, Bricks, Beaver Builder, Oxygen,
Breakdance, WP Bakery and Gutenberg. The point-and-click Magic Bar and full-page
tests work with any theme or builder, because they operate on the rendered page.

### What counts as a conversion?

A visit to the page you choose, such as your thank-you page, or a click on an element
you choose with a CSS selector, such as a buy button. To count form submissions 
or WooCommerce purchases, choose the page visitors land on afterwards (the form’s
thank-you page, or the WooCommerce order-received page).

### How long is heatmap data kept?

3 days by default. You can change the number of days under **AB Split Test Lite  
Settings  Heatmaps**.

### Will it slow down my site or fight my caching plugin?

The tracking script is small, loads from your own site, and makes no requests to
third-party servers. It is automatically excluded from minification and delay in
WP Rocket, NitroPack, LiteSpeed Cache, SiteGround Optimizer and other optimization
plugins.

Tests work on cached pages, including host-level page caches, because each visitor
is assigned to the original or the variation in their own browser and visits and
conversions are recorded from the browser too. You don’t need to exclude tested 
pages from your cache.

### Does it use cookies? Is it GDPR friendly?

AB Split Test uses first-party cookies and browser storage on your own domain to
remember which version a visitor saw. It does not fingerprint devices and sends 
nothing to a third party.

If you need consent first, turn on **Wait for cookie consent**. Tests still run,
but nothing is stored or sent until the visitor agrees. It works with Cookiebot,
CookieConsent, Usercentrics (including Termageddon) and any consent plugin that 
supports the WP Consent API (such as Complianz or CookieYes), or call `setAbstApprovalStatus(
true)` from your own banner when the visitor accepts; it is remembered on later 
pages. Call `setAbstApprovalStatus(false)` when consent is withdrawn; this clears
tracking identifiers and pending history. Until consent is confirmed, assignments
stay in page memory. As with any analytics tool, check your own consent requirements.

### What happens to my data if I delete the plugin?

Your tests and results are kept by default, so nothing is lost by accident. Developers
who want a full clean-up can set the `abst_delete_data_on_uninstall` option to `
1` before deleting the plugin.

### Do I need to know how to code?

No. The Magic Bar lets you build tests by clicking elements on the page. Developers
who want more control get the REST API and MCP tools.

### Where can I get help or report a bug?

Use the [WordPress.org support forum](https://wordpress.org/support/plugin/ab-split-test-lite/)
or visit [absplittest.com](https://absplittest.com) for documentation.

## Reviews

There are no reviews for this plugin.

## Contributors & Developers

“AB Split Test Lite” is open source software. The following people have contributed
to this plugin.

Contributors

 *   [ Tom ](https://profiles.wordpress.org/tomsitespot/)

[Translate “AB Split Test Lite” into your language.](https://translate.wordpress.org/projects/wp-plugins/ab-split-test-lite)

### Interested in development?

[Browse the code](https://plugins.trac.wordpress.org/browser/ab-split-test-lite/),
check out the [SVN repository](https://plugins.svn.wordpress.org/ab-split-test-lite/),
or subscribe to the [development log](https://plugins.trac.wordpress.org/log/ab-split-test-lite/)
by [RSS](https://plugins.trac.wordpress.org/log/ab-split-test-lite/?limit=100&mode=stop_on_copy&format=rss).

## Changelog

#### 1.0.0

 * First WordPress.org release of AB Split Test Lite, built on the AB Split Test
   engine used since 2019.
 * Self-hosted point-and-click, full-page and on-page A/B tests, each comparing 
   the original with one variation, with no limit on the number of tests.
 * MCP tools (via the WordPress Abilities API) and REST API for AI-agent and programmatic
   control.
 * Click heatmaps and scroll maps on every page, with configurable data retention.
 * Native page-builder controls for Elementor, Bricks, Beaver Builder, Oxygen, Breakdance,
   WP Bakery and Gutenberg.
 * Bayesian statistical analysis with device breakdown.
 * An email to the site admin when a running test finds a winner. The test keeps
   running until you end it.

## Meta

 *  Version **1.0.0**
 *  Last updated **23 hours ago**
 *  Active installations **Fewer than 10**
 *  WordPress version ** 6.9 or higher **
 *  Tested up to **7.1.5**
 *  PHP version ** 7.4 or higher **
 * Tags
 * [a b testing](https://wordpress.org/plugins/tags/a-b-testing/)[conversion optimization](https://wordpress.org/plugins/tags/conversion-optimization/)
   [heatmap](https://wordpress.org/plugins/tags/heatmap/)[mcp](https://wordpress.org/plugins/tags/mcp/)
   [split testing](https://wordpress.org/plugins/tags/split-testing/)
 *  [Advanced View](https://wordpress.org/plugins/ab-split-test-lite/advanced/)

## Ratings

No reviews have been submitted yet.

[Your review](https://wordpress.org/support/plugin/ab-split-test-lite/reviews/#new-post)

[See all reviews](https://wordpress.org/support/plugin/ab-split-test-lite/reviews/)

## Contributors

 *   [ Tom ](https://profiles.wordpress.org/tomsitespot/)

## Support

Got something to say? Need help?

 [View support forum](https://wordpress.org/support/plugin/ab-split-test-lite/)