Split Testing Landing Pages With Redirects in WordPress
A rotating redirect is the simplest possible split test: two real pages, one entry link, an even split. Here is how to run one without hurting your SEO.
On this page
You do not need a testing platform to compare two landing pages. You need two real URLs and one link that alternates between them.
A rotating redirect does exactly that: visitors clicking your entry link get sent to variant A or variant B, roughly evenly, and each variant is an ordinary page you can measure like any other. No script, no third party tag, nothing for a blocker to strip.
It is the simplest split test that works, and the setup takes about ten minutes.
The Setup
Build the two variants as normal pages: /offer-a/ and /offer-b/. Make them genuinely different in one respect you care about, not twenty, or you will not know what caused any difference you see.
Create one entry URL, something like /go/offer/, and set it to rotate between the two destinations. Then use that entry URL everywhere: in your ads, your newsletter, your social posts, your sidebar.
DevDome Redirect Manager handles the rotation and counts clicks on the rule, which gives you the top of the funnel number for free. It is fully free with no rule limit, so a test does not cost you anything but the time.
Use 302, Always
This is the one setting that determines whether your test is a test.
A 301 tells the browser the move is permanent, and browsers cache that aggressively. A visitor who sees variant A on Monday keeps being sent to variant A for as long as their browser remembers, so your split stops splitting and your data quietly becomes a study of who happened to arrive first.
A 302 says temporary, which is exactly what a test is. The browser re-checks each time and the rotation keeps working.
Confirm with the redirect checker after you set it up. Seeing a 301 where you expected a 302 is a common and silent failure.
Keep Search Engines Out of It
Google’s testing guidance is clear that testing is normal, with conditions. Three things to get right.
Do not rotate crawlers. A crawler bounced between two variants indexes an unstable picture of your site. Serve verified search crawlers one consistent version.
Set a canonical. Each variant should carry a canonical tag pointing at the URL you want indexed, usually the original page. This tells search engines the variants are versions of one thing rather than duplicate content competing with each other.
Keep it temporary. A test is a test. A permanent rotating split with no end date starts to look like something else, and it stops being possible to reason about which URL should rank.
Decide the Rules Before You Start
Write down two things before the first visitor arrives.
What counts as winning. Signups, purchases, outbound clicks, contact forms. One metric, chosen in advance. Deciding afterwards means choosing whichever number happened to favour the variant you liked, and everyone does this unless they wrote it down first.
How many visitors you need. Any online sample size calculator will give you a figure from your current conversion rate and the improvement you would care about. The number is usually larger than people expect, which is itself useful information: if the test needs 40,000 visitors and you get 400 a week, this is not a test you can run.
Do Not Stop Early
The single biggest source of wrong conclusions in split testing.
Early results swing wildly. A variant that is 40 percent ahead after 80 visitors will very often be behind after 800, and if you stop at the moment it looked best you have measured noise and called it a finding.
Set the sample size, run to it, then look. Checking daily is fine as long as checking does not become deciding.
Reading the Result
You have two numbers per variant: how many arrived, and how many converted.
The redirect rule gives you the entry clicks. Your analytics gives you the pageviews and the conversions per variant, since each is a separate URL. DevDome Analytics captures outbound clicks server-side, which matters if your conversion is a click out to a merchant, because that is exactly the event client-side tracking loses most of.
Compare conversion rate, not raw conversions, since the split is rarely perfectly even. A rotation that lands 520 visitors on A and 480 on B is normal, and raw totals will mislead you by exactly that margin.
Be careful with traffic that arrived from somewhere unusual during the test window. A mention on a large site, a newsletter send, or a burst of social traffic all bring an audience that behaves differently from your baseline, and if that burst happened to land mostly on one variant it is now sitting inside your result. If you can, look at the two variants within a single traffic source rather than across all of them.
And take a look at bot share while you are there. If one variant received an unusual amount of automated traffic, its rate is distorted, and a test decided on bot behaviour is worse than no test.
End It Properly
When you have a winner, finish the job.
Point the entry URL directly at the winning variant, and 301 the losing variant to the winner so any links that reached it are not wasted. Remove the canonical you added. Note what you tested, what you found and how many visitors it took, because in six months you will not remember and someone will propose the same test again.
If the result is inconclusive, that is a real answer too. It means the difference you tested does not matter enough to measure, and you can stop thinking about it.
Disclosure: DevDome publishes this blog and makes the products it mentions. We describe what they do rather than promise results.
Key takeaways
- A rotating redirect splits traffic between two real URLs with no testing script.
- Always use 302, never 301, or browsers will lock visitors to one variant.
- Exclude search crawlers and set a canonical so you are not testing your SEO.
- Decide the sample size and the winning metric before you start, not after.
- Stop the test when you decide, and redirect the loser to the winner.
Sources
- Google Search Central: website testing and Google Search — Google's guidance on running A/B tests without SEO problems
- MDN: HTTP 302 Found — why a temporary redirect is the correct status for a test
Links last checked September 2, 2026.
Frequently asked questions
Will split testing hurt my SEO?
Not if you set it up properly. Serve search crawlers a consistent version rather than rotating them, point a canonical from the variant to the main URL, and keep the test time-limited. Google has long said testing is normal site behaviour; what causes problems is leaving a permanent split running or showing crawlers something different from visitors.
How long should a test run?
Long enough to cover your normal weekly cycle, so at minimum a full week, and long enough to reach the sample size you decided on beforehand. Stopping the moment one variant pulls ahead is the classic mistake: early leads reverse constantly and you will convince yourself of something untrue.
Why 302 and not 301?
A 301 is cached hard by browsers, so a returning visitor keeps landing on whichever variant they saw first and the split stops being a split. It also tells search engines the move is permanent, which is the opposite of what a test means. A 302 keeps the browser asking each time.
Can I test more than two variants?
Technically yes, practically be careful. Every extra variant divides your traffic further, so a three-way test needs substantially more visitors to reach the same confidence. Most small sites do not have the traffic for more than two, and a badly powered three-way test is worse than a clean two-way one.