Short version: An A/B test compares two versions of the same page, message, design, ad, or flow to learn which one performs better against a defined goal.
The original version is usually the control and the changed version is the variation. Visitors are split between them, performance is measured, and the result is evaluated against a metric such as clicks, leads, purchases, signups, or conversion rate. The value of an A/B test is not that it makes a team guess less loudly; it creates evidence for a specific decision.
A useful A/B test needs one clear hypothesis, enough traffic or conversions to read the result responsibly, and a practical view of impact. A statistically significant lift that is too small to matter commercially may not be worth implementing. Public SEO-sensitive tests also need care: avoid cloaking, keep the user experience consistent with search expectations, and use canonical or temporary redirects when separate URLs are involved.
Related terms: CRO, conversion rate, and CTA.