A/B test.

In an A/B test, you pit two versions against each other: two ad images, two headlines, or two buttons. One group of people sees Version A, the other sees Version B, and in the end, you determine which one generates more inquiries or purchases. For a business, this replaces gut feelings with measurable data—provided enough people participate in the test.

How an A/B Test Works

You change exactly one thing and leave the rest the same. According to Meta, an A/B test in Ads Manager compares two ad variations that differ in one variable, such as image, text, audience, or placement. According to Meta’s help center, Meta shows each variation to only a portion of the target audience so that no one sees both, and recommends the same budget for both variations.

In Google Ads, this tool is called a custom experiment. According to Google, the experiment uses a portion of the traffic and budget from the original campaign; for a clean comparison, Google recommends splitting them 50-50. The process is similar on your own website: According to Google Search Central, you can create variations under separate URLs or switch between them on the same page using JavaScript.

What Your Business Should Keep in Mind

Write down your expectations beforehand. Meta recommends formulating a hypothesis before choosing a variable. A statement like “A photo of the actual team generates more inquiries than a stock image” is enough to ensure that you’ll know what you’ve learned afterward.

Don't run tests by manually turning ads on and off. According to Meta, this leads to inefficient delivery, unreliable results, and overlapping audiences. Use the built-in testing features from Meta and Google.

Protect your Google ranking while testing your website. Google Search Central recommends using a canonical tag pointing to the original page for variants with their own URLs, 302 redirects instead of 301 redirects, and no cloaking. According to Google, alternative URLs and test code should be removed quickly after the test is complete.

Plan for enough visitors. According to Google, the required test duration depends on your conversion rates and the number of page views. In projects, we often see that landing pages don’t have enough visitors to detect subtle differences. In that case, it’s better to test major changes—such as a different offer—rather than just a different button color.

Frequently Asked Questions About A/B Testing

What should I test first? The element that has the greatest impact on the decision—often the headline, the offer, or the call to action. Google itself cites the text on a button—such as “Add to Cart” versus “Buy Now”—as an example of a small change with a surprising effect.

What's the difference between this and a multivariate test? According to Google, a multivariate test evaluates multiple changes at the same time, as well as how they interact with each other. For small websites, a simple A/B test is usually the more realistic approach.

Will an A/B test hurt my Google ranking? Not if you follow the guidelines from Google Search Central: no cloaking, use a canonical tag pointing to the original page, use temporary redirects, and keep the test running no longer than necessary.

Before you run a test, it's worth taking a look at where visitors get stuck on your page. The Design Check will show you this, and you can find tips on how to build a testable landing page in " Create a Landing Page."