Split URL Testing: What It Is, How It Differs From A/B Testing, and How to Run One

The short answer: As of October 8, 2026, split URL testing (also called a redirect test) is an experiment where each version of a page lives at its own URL. Visitors who land on the control URL are randomly kept there or redirected to a variant URL, and the tool compares conversions. It is the right method when the change is too big to edit in place: a redesign, a new template, a rebuilt checkout or a page built in a different tool. Google's guidance for keeping these tests safe in search: put rel="canonical" on variants pointing to the original, use a 302 redirect (or a JavaScript redirect), never cloak, and run the test only as long as necessary.

TL;DR

• A split URL test is an A/B test where the versions are separate pages, not edits to one page.
• Use it for big swings. Use an on-page A/B test for one headline, image or button.
• Multivariate testing answers a different question (which combination of elements wins) and needs far more traffic.
• For SEO, follow Google's four rules: canonical, temporary redirect, no cloaking, short duration.
• Most broken split URL tests come from mismatched URLs (trailing slashes, www), lost UTMs and double-counted page views.

What split URL testing is

In a split URL test you build two or more complete pages:

• Control (A): your current page, at its current URL, for example /pricing.
• Variant (B, C, and so on): a new version at a different URL, for example /pricing-b.

The testing tool sits on the control. When a visitor arrives, it assigns them to a version. Visitors assigned to the control stay put; visitors assigned to a variant are redirected to that variant's URL. Every version reports the same goal (a signup, a form submission, a purchase, a visit to a thank-you page), and the tool compares conversion rates.

Google describes the same method in its testing guidance: create multiple versions of a page, each with its own URL, and when users try to access the original URL, redirect some of them to each variation URL (Google Search Central, Minimize A/B testing impact in Google Search).

Split URL test vs A/B test vs multivariate test

Two points the table hides:

1. A split URL test is a kind of A/B test. The statistics are the same: compare conversion rates between groups and check whether the difference is bigger than chance. Only the delivery differs.
2. Multivariate testing is a different question. Google's definition: testing more than one type of change at a time, looking at the impact of each change and the synergies between them. Every combination becomes its own cell, so traffic requirements multiply. Humblytics does not offer multivariate testing. You can run a split URL test with more than two variants, which covers most "which of these three pages wins" questions.

When to use split URL testing

Pick split URL testing when the variant is a different page, not a tweak to the same one:

• Redesigns. A new layout, new information hierarchy or new visual system. Rebuilding that as in-browser edits is fragile and slow.
• New templates. A new product page, pricing page or blog template you want to prove before rolling it out site-wide.
• Different page builders or stacks. The control is in Webflow and the variant is a Framer page, or the variant runs on a new Next.js build. A redirect test does not care how each page was made, as long as both carry the same tracking.
• Landing page and checkout rebuilds. Single-step vs multi-step checkout, a long-form vs short-form landing page, a new signup flow.
• Backend or performance changes. A page with a different data source, a lighter build or a new form handler, where the difference is not visible in a page editor.

Stick with an on-page test when you are changing one element. It is faster to build and you avoid redirect pitfalls entirely. Whatever you test, start with a written hypothesis: how to write an A/B test hypothesis.

How split URL testing works technically

The redirect

The control URL loads the testing script, the script assigns the visitor, and assigned visitors are sent to the variant. That redirect can be server-side (an HTTP 302) or client-side (JavaScript). Google says JavaScript-based redirects are fine for tests.

Flicker

The classic problem with client-side redirects is a flash of the control page before the variant loads. It looks broken and can bias results, because people who see a flash behave differently. In Humblytics, variant redirects happen before paint, so visitors do not see the original content first (Humblytics Docs, How split testing works under the hood).

Analytics

A redirect means a visitor technically touched two pages: the control, then the variant. If your analytics counts both, the control's page views inflate and the variant's referrer can show up wrong. Humblytics suppresses events for the original page when it redirects, so a redirected visitor is counted once, on the variant.

SEO: what Google actually says

Google's page on testing lists four best practices (Google Search Central, last updated December 10, 2025):

1. Don't cloak. Do not show one set of URLs to Googlebot and a different set to people. Google says cloaking is against its spam policies whether you are running a test or not, and that can get a site demoted or removed.
2. Use rel="canonical" on variant URLs, pointing to the original. Google recommends this over noindex, because you want the variants grouped with the original, not the original dropped.
3. Use 302 redirects, not 301. A 302 tells search engines the redirect is temporary, so they keep the original URL in the index instead of replacing it with the test page. JavaScript redirects are also acceptable.
4. Run the experiment only as long as necessary. When the test ends, ship the winner and remove the alternate URLs and test code. Google says it may treat an unnecessarily long experiment as an attempt to deceive search engines, especially when one variant is served to a large share of users.

Google also notes that small changes such as button color or call-to-action text often have little or no impact on a page's search snippet or ranking.

How this maps to Humblytics: variant pages automatically point to the control with canonical tags, and the testing logic is skipped for known bots and crawlers, so crawlers see the control, which is the same URL the variants canonicalize to. Duration is on you and on the test settings: tests can stop automatically after a set duration (30 days by default), and you can stop one manually at any time.

How to set up a split URL test in Humblytics

You need the Humblytics script installed once on your site. It covers every page, including new variant pages, so no extra snippet is needed per variant.

1. Build the variant page. Duplicate the control in your site builder, make your change and publish it at a short, readable slug like /pricing-b. Walkthroughs for specific builders: split testing in Webflow and split testing in Framer. The Humblytics docs also cover WordPress, Shopify and Wix. For custom code, deploy the variant template at its own path and keep the script in the <head>.

2. Start a new experiment. In Humblytics, open Split Testing in the left menu and click Start New Experiment.

3. Choose the goal. Conversion goals: form submission, click, page view or purchase event. Page goal: Reach Destination Page, which accepts an internal path like /thank-you or a full external URL like a separate checkout domain. External destinations count only when the visitor goes there directly from your site.

4. Name the experiment. Be specific, for example "Pricing page: current vs rebuilt template". Add your hypothesis in the notes.

5. Add the URLs. Paste the Control URL and one or more Test Variant URLs. Use Add Variant for a third or fourth page; traffic splits evenly across them by default.

6. Set visitor assignment.

• Allow Visitor Overlap: assignment per session, so a returning visitor may see a different version. Fine for short, high-traffic tests.
• No Visitor Overlap: assignment per user, so they always see the same version. Use it for redesigns, long funnels and anything brand-sensitive, which describes most split URL tests.

7. Optionally restrict the audience. Under Audience & location, include only certain device types, traffic sources (organic, paid, direct, referral) or countries. Leave an axis empty to include everyone.

8. Review and launch. Click Start Experiment. Traffic starts splitting immediately.

9. Read the results. Humblytics shows conversion rate, relative lift and confidence per variant, calculated with a two-proportion z-test. The default confidence threshold shown is 95%. Reaching it does not end the test; the test runs until you stop it or until its set duration ends. Promoting the winner is your call.

More on the product side: Humblytics A/B testing.

Pitfalls that break split URL tests

Trailing slashes and www

URL matching in Humblytics is exact, including the trailing slash. /pricing-b and /pricing-b/ are different URLs. Many hosts redirect one to the other, and a variant entered in the wrong form shows zero traffic. The fix: open each URL in a browser, let any redirect finish, and copy the final URL from the address bar. Do the same for www vs non-www.

Redirects that strip UTMs and referrers

Any redirect is a chance to lose the query string. If UTM parameters do not survive the hop from control to variant, paid and email visits on the variant look like direct traffic, and your channel breakdown skews by variant. Before launch, open the control URL with a test UTM like ?utm_source=test, get assigned to the variant (use a private window and retry), and check that the parameters are still in the variant URL and show up in your reports. Unexplained growth in direct traffic on one variant is the tell; see what direct traffic is and why it grows.

Double counting

If you run a second analytics tool alongside your testing tool, it may count a page view on the control and another on the variant for the same person. Compare totals across tools before trusting either. Within Humblytics, the original page's events are suppressed on redirect.

Changing more than you meant to

A redesign changes many things at once. That is the point, but it means a win tells you the new page is better, not which part made it better. If you need to know why, follow up with smaller on-page tests.

Sample size and duration

Split URL tests usually test big changes, which helps: bigger effects need fewer visitors to detect. They still need enough conversions per variant, and splitting across three or four variants divides your traffic further. Run full weekly cycles so weekday and weekend behavior both count. If your site is small, read A/B testing with low traffic before you add a third variant.

Leaving the test running

When a winner is clear, publish it at the original URL, remove the variant pages or redirect them permanently, and stop the experiment. That is both good hygiene and what Google asks for.

Run your first split URL test

Humblytics runs analytics, heatmaps, funnels and split testing from one script, and matches Meta and Google Ads spend to Stripe revenue. Build your variant in the tool you already use, paste two URLs and launch. See Humblytics A/B testing or start a 14-day trial.

Sources and freshness

• Google Search Central, Minimize A/B testing impact in Google Search, last updated December 10, 2025, retrieved October 8, 2026. Supports the definitions of A/B and multivariate testing, the redirect method, and the four best practices: no cloaking, rel="canonical", 302 or JavaScript redirects, and limited test duration.
• Humblytics Docs, How to set up a split test, retrieved October 8, 2026. Supports the setup steps, goal types, visitor assignment options, audience restrictions and exact URL matching including trailing slashes.
• Humblytics Docs, How Humblytics split testing works under the hood, retrieved October 8, 2026. Supports redirects before paint, event suppression on redirect, bot handling, automatic canonical tags and the two-proportion z-test.
• Humblytics Docs, Creating A/B page variants for split testing, retrieved October 8, 2026. Supports building variants in Webflow, Framer, WordPress, Shopify and Wix with one global script.
• Humblytics product facts (duration-based auto-stop, 30-day default duration, 95% default confidence threshold) from Humblytics' published product information, current as of October 8, 2026.

Testing tools and Google's documentation change. Check the linked pages before relying on a specific setting.