Landing page A/B testing: what to test, which tools and how long

Landing page builders with built-in A/B testing

Most dedicated landing page builders sell testing on their higher plans. On October 11, 2026, Unbounce included A/B testing from its Experiment plan at $149/month, or $112/month billed annually, with unlimited tests and variants, manual traffic allocation and up to 30,000 visitors a month; its Starter and Build plans don't include A/B testing. Unbounce's Smart Traffic, which routes each visitor to the variant they're most likely to convert on, starts on Optimize at $249/month ($187 billed annually). Unbounce counts each visitor once a month toward plan limits and bills overages.

Instapage puts server-side A/B testing on its Optimize plan at $199/month, or $159/month billed annually, for 30,000 unique monthly visitors; its Create plan at $99/month ($79 annually) doesn't list A/B testing. Leadpages includes manual A/B testing from its Grow plan, with up to 6 variants per page, but its pricing page shows two different sets of prices, so check it directly. Carrd's Pro feature list doesn't include A/B testing, though Pro Standard and above allow embeds, which is how a third-party script gets onto the page.

Built-in testing has one real advantage: the variants live in the same builder as the page, so there's no extra script. The trade-offs are that pricing rises with visitors and that testing covers only pages built in that tool. We make Otter, so read the next section with that in mind, and see our A/B testing tools comparison for more options.

How Otter runs landing page split testing on any builder

Otter works wherever you can add a script to the page head. Webflow, Framer, Wix, Squarespace and ClickFunnels each have a custom code or header tracking field, WordPress has Otter's plugin, and hand-coded pages take the snippet directly. One tool can then test landing pages across builders, and the price doesn't change with traffic: Starter is $39/month for one site with unlimited visitors and tests. The Webflow A/B testing, Framer A/B testing and HubSpot A/B testing guides cover platform details.

For changes on one page, open it in Otter's visual editor and edit the headline, copy, images (by URL), styles or the content around the form, with mobile and tablet previews. For a complete redesign, run a split URL test between two published pages. Split URL redirects happen in the browser, and with relay parameters on, the visitor's query string and hash carry over, so UTM tags and ad click IDs reach the variant page and your analytics still credit the campaign.

Paid landing pages often deserve their own audience. Otter can target a test by UTM source, medium or campaign, traffic source (such as paid search, social or email), device, country, new or returning visitors and URL parameters, combined with AND/OR rules. That lets you run a message-match test only for visitors from one campaign while everyone else sees the original.

What to test first on a landing page

Message match comes first. If an ad promises "payroll in 10 minutes", the headline should say the same thing, and testing a headline that mirrors the ad against a generic one is the classic first test. Next, test the form. Fewer fields often mean more submissions but sometimes less qualified leads, so decide in advance whether you're optimizing for form fills or for leads that sales accepts.

After that, test the call-to-action wording and position, the proof near the form, such as customer logos, a testimonial or a review score, and the offer itself: a free trial against a demo, or a guide against a checklist. Test one idea per variant so you know what caused the result. When you have several versions of the same idea, run an A/B/n test; Otter raises the significance bar automatically as you add variants, so extra variants don't inflate false wins. To choose between ideas, score them with the test prioritization calculator.

Measuring leads: goals for landing page tests

The goal defines the test, so pick it before you build anything. Otter has no form-submit goal type, so landing page tests usually use one of four goals. A pageview goal on the thank-you URL counts completed submissions and is the most reliable choice when the form redirects. A click goal on the submit button counts attempts, including ones that fail validation. A custom event, sent with optimo.track() when your form code confirms success, is exact but needs a developer. A GTM dataLayer event works if your tag setup already pushes a form event.

If GA4 already records leads, for example as a generate_lead event, Otter can import that GA4 event as a goal once you connect GA4 and register two custom dimensions. Results then use GA4's own user counts and arrive after GA4's processing delay. For pages that sell directly, track revenue with optimo.revenue() and judge the test on revenue per visitor.

Landing page split testing pitfalls

Campaign changes mid-test are the biggest risk. If someone raises a budget, changes targeting or launches a new ad halfway through, the traffic mix shifts under the test. Annotate those changes on the results chart and, if they're large, restart. Run tests in whole weeks, since weekday and weekend visitors behave differently, and don't stop a frequentist test the moment it crosses your threshold. Plan the sample size and wait for it, or use Otter's Bayesian mode, which reports each variant's chance to beat the original.

Check the traffic split as well. If a 50/50 test shows 5,400 visitors on one side and 4,600 on the other, something is losing visitors, often a redirect that fails in some browsers or a variant page that loads slowly. Otter's Trust Check runs a sample ratio mismatch test, and the free sample ratio mismatch checker does the same for results from any tool. Finally, a split URL test is a browser redirect: Otter keeps the page hidden until it decides, with a 3-second failsafe, so keep variant pages fast.

Landing page A/B testing and SEO

Many paid landing pages are kept out of search on purpose, and then SEO doesn't constrain the test. For pages that rank, follow Google's guidance on website testing: don't cloak, use a 302 or JavaScript redirect for split URL tests, put a rel="canonical" tag on the variant page pointing to the original rather than noindexing it, and end the test once you have a result.

Visual-editor tests keep one URL and change content with JavaScript, which Google lists as a valid approach, and small changes to buttons or copy often have little or no effect on rankings. Otter doesn't assign known bots to a variant, so crawlers index the original. After a test, publish the winner as the only version and delete or redirect the losing page.