FROM THE BLOG

AB Split Test v2.6.5: Custom Audiences for Better Referrer Insights

The latest version of AB Split Test adds Custom Audiences: reusable visitor segments you can attach to any test so you can see how different traffic sources behave inside the same experiment.

Watch the walkthrough video below, then try it on your next test. This is especially useful when one traffic source loves a variation and another source clearly does not.

What Are Custom Audiences?

Custom Audiences let you save visitor segments once, then reuse them across tests. Instead of only asking “which page won?”, you can ask better questions:

  • Did YouTube visitors prefer the same variation as ChatGPT visitors?
  • Are Meta Ads visitors behaving differently from Reddit visitors?
  • Does the overall winner still win for high-intent traffic?
  • Are AI referral sources converting differently from search or social traffic?

Why do I want this?

An A/B test can have a clear overall winner and still hide something important. A new headline might win across all visitors, but lose badly for ChatGPT traffic. Or a pricing-page variation might be average overall but perform beautifully for YouTube visitors.

Custom Audiences make that visible. You still keep one main result set for the test, but you also get filtered result views for the audiences you care about.

What You Can Segment By

  • URL query rules, like utm_source=google, gclid, or fbclid
  • Session referrer rules, like youtube., chatgpt.com, claude.ai, or reddit.
  • Browser language, browser, and operating system
  • Day and time windows in your WordPress site timezone
  • Cookie rules for returning users, consent states, campaigns, or custom site segments

There are also presets for common sources, including Google Ads, Meta Ads, YouTube, LinkedIn, Reddit, Pinterest, ChatGPT, Claude, Gemini, Perplexity, Copilot, Grok, DeepSeek, Mistral, Poe, and You.com.

How It Works

  • Create reusable audiences under Split Tests → Audiences.
  • Attach the audiences you want inside a test’s visitor segmentation settings.
  • Run the test as normal. All eligible visitors still count toward the main result.
  • Open the results and switch the audience filter to see how each segment performed.
A website admin dashboard shows audience settings for A/B split testing, listing platforms like Facebook, Reddit, and ChatGPT, with options to edit or delete each audience rule.

Audience subsets can overlap, so their totals do not need to add up to the overall test total. That is intentional. The goal is not to replace the main result, but to reveal the pockets of traffic where your winner is stronger, weaker, or completely reversed.

A Better Way To Read Referrer Tests

This update makes referrer analysis much more useful. You can now run one test and quickly see if YouTube, ChatGPT, Claude, Reddit, Meta, or Perplexity visitors are responding differently.

That means fewer duplicate tests, fewer messy spreadsheets, and a much clearer answer to the real CRO question: who is this change actually working for?

A dropdown menu labeled "Audience" shows options like Reddit, YouTube, and ChatGPT visitors. Below, a chart displays conversion data and a line graph.

Get The Update

Custom Audiences are available in the latest version of AB Split Test. Update the plugin, watch the video above, and try adding a few referrer audiences to your next experiment.