In most conversations there's no difference: "A/B testing" and "split testing" both mean randomly dividing your audience between versions and comparing what they do. When people do draw a distinction, "split testing" usually means a split URL test, where each version is a separate page and visitors are redirected to it.
Where the two names come from
"A/B" names the versions: A and B. "Split" names the method: you split the traffic. Email marketers and ad platforms have long said "split test"; website optimization tools popularized "A/B test". Meta, for instance, used "split testing" in its ad tools for years before moving to "A/B test" in its interface.
The one real distinction: in-page vs split URL
In-page A/B test. Both versions live at the same address. A script or the server swaps elements (headline, image, layout) depending on which group the visitor is in. Good for changes to parts of a page.
Split URL test. Version B is a separate page, say /product-new instead of /product. Visitors assigned to B are redirected. Good for complete redesigns, different checkout flows or pages built in a different tool.
| In-page A/B | Split URL | |
|---|---|---|
| Same URL | Yes | No, redirect to B |
| Best for | Element changes | Whole-page or flow changes |
| Main risk | Flicker, script weight | Visitors lost in the redirect, SEO handling |
| Setup | Visual editor or code | Build a second page |
Watch the redirect
In a split URL test, some visitors leave during the redirect, especially on slow mobile connections. They count as assigned to B but never see it, or never get tracked at all. The result is a lopsided split, and the people who went missing are likely to be the impatient ones, who convert differently. Check every split URL test with the SRM checker.
Keeping search engines happy
Search engines are generally fine with testing as long as you aren't trying to fool them. For split URL tests:
- use a temporary (302) redirect, not a permanent (301) one,
- add
rel="canonical"on the variant page pointing to the original, - show crawlers the same thing users would see,
- finish the test and remove the variant when you're done.
Bottom line
Use whichever term your team prefers. What matters is the method underneath: random assignment, both versions at the same time, a pre-planned sample, and an honest analysis. For the full list of designs, see types of A/B testing.
Common questions
Is split testing the same as A/B testing?
In everyday use, yes. Both mean splitting an audience at random between versions and comparing results. Some tools use "split test" specifically for split URL tests, where each version is a separate page.
Are split URL tests bad for SEO?
Not if they're done properly. Use temporary (302) redirects rather than permanent ones, add a canonical tag on the variant page pointing to the original, don't show search engines a different version than users, and end the test when it's done.