Test a new page without touching the ad link

The situation is familiar. An ad set has found its audience, the cost per sale is where you want it, and you have a new page you think will convert better. Pointing the ad at a new address feels like the riskiest edit you could make to something that works.

We wrote this from vendor documentation, Google's help pages and media buyers' own reports. We do not run ad accounts to test these claims, and where the evidence is a practitioner report, the page says so.

Same-URL testing depends on server-side split testing: the decision is made before the page leaves the server. The alternative is a client-side script that loads the original page and then rewrites it in the browser. On a slow phone connection the visitor can see the original for a moment before the variant replaces it, and the page jumps.

Leadpages describes its own approach in those terms. Its A/B testing page says it assigns variants on the server before the page renders, and claims no flicker, no layout shift and no original version flashing past. That is the vendor describing its product, but the mechanism behind it is sound: a page that arrives finished has nothing to swap.

Server-side assignment alone does not make a test same-URL. A builder can pick the variant on the server and still publish each variant at its own path, and the Leadpages pages we read do not say whether its variants share one public address. Ask any vendor that question directly before you plan around it.

One naming clash to clear up: server-side tracking is a different subject. It is about how conversions get reported back to the ad platform, which our guide to server-side tracking vs the browser pixel covers.

The working rule from the Meta thread is simple: keep the landing page URL in the ad and change what sits behind it. There are three ways to do that.

1. A page tool that rotates variants under one address. You build B next to A, set the weights and leave the campaign alone. Shifting all traffic to the winner is a setting in the page tool.
2. A redirect you control. The ad points at a path on your own domain that forwards to whichever page you want live. Changing the target is a server edit. A redirect adds a hop, and the visitor ends up on a different address from the one in the ad.
3. Rotation in your ad tracker. The campaign link points at the tracker, which sends visitors to one of several landers. The ad link never changes, though each lander keeps its own address.

A split URL test is still a fine choice when nothing is at stake: a brand-new campaign that has not exited learning, organic or email traffic, or a test you are happy to run on a fresh ad set.

Whichever route you take, the test still needs enough conversions per variant before you call it. Fewer variants reach that point sooner, and the break-even ROAS and CPA calculator tells you what a winning variant has to beat.

Vendors that sell same-URL testing say the ad platform has no new destination to review, so it has no reason to restart learning. That is a sales argument, and no independent measurement we found shows how often a URL swap on its own resets learning compared with the other edits buyers tend to make at the same time.

A fixed address also leaves the experience behind it open to change. A variant that loads slower or sends visitors bouncing changes what the platform learns from your traffic, whatever the URL says. And a test with too few conversions per variant gives you a coin flip, under one address or two.