How to Use Framer to A/B Test a Landing Page Headline

The short answer: if the landing page converts on the first visit, with a signup or a demo click, Framer's built-in test settles a headline argument, and it needs a Pro site at $30 a month billed yearly plus the Convert add-on at $50 per 500,000 events (pricing page, checked on 26 October 2026). The steps: put a tracking ID on the call to action, right-click the page and choose New A/B Test, write the new headline on the variant or have an agent draft it on a branch for you to review, set the conversion step in Analytics, click Start, publish and deploy, and leave the variants alone until Framer calls a winner or the test is inconclusive.

Disclosure: I may earn a commission through some links on this site. It does not change what this page says.

The job: one campaign lands on one page and the team disagrees about the headline. Framer shows each visitor one of up to five variants, splits traffic evenly and reports conversion rate, lift against the control and each variant's probability of being ahead, per its A/B testing guide. Plans and limits are in the Framer glossary entry.

Three checks before you start:

1. The add-on bill. Convert counts every page view, every click on a tracked link and every tracked form submission across the site. For example, a site with 120,000 page views and 3,000 tracked clicks a month generates 123,000 events, inside one 500,000 block, so the test adds $50 a month on yearly billing or $60 on monthly, per the pricing page.
2. The tracking you already have. If the button or form already carries a tracking ID, its clicks sit in the Tracking section of the Analytics Overview, which lists the top 20 events, per the tracking guide. That history is your baseline conversion rate.
3. The window the test can see. Framer measures conversions within the same day, because its visitor identifier resets daily, per the A/B testing guide. If buyers come back a week later to book a call, plan a second check in your CRM.

One more thing decides how you set this up in Europe. Framer runs these tests without cookies: it builds an anonymous identifier from the visitor's IP address and browser information, and says that can remove the need for a cookie banner depending on local law. The EDPB's guidelines on Article 5(3) of the ePrivacy Directive say that collecting information through an IP address or a browser's user agent, as fingerprinting does, can fall under the same consent rule as cookies. Ask whoever owns privacy before a banner comes down on the strength of the test.

What you need

• A Pro or Enterprise site with the Convert add-on, which the guide makes a condition for A/B tests and funnels.
• A page with one clear conversion: a link click, a form submission or a visit to a thank-you page.
• For the agent step, an MCP client such as Claude, Codex or Cursor connected to Framer, set up as the MCP guide describes.

Step 1: Tag the conversion

Select the call-to-action link, or the form, on the canvas and add a tracking ID during setup. The tracking guide asks for a unique ID so the results do not mix with other events, and the same ID must sit on that element in every variant.

Step 2: Create the variant

Right-click the page and choose New A/B Test. Framer copies the page into a new variant, which you can rename; a page takes up to five. Give each a name that says what changed, such as "Headline: outcome" and "Headline: price". Change only the headline, so the result tells you something about the headline.

Step 3: Let an agent draft the headline on a branch

Connect Framer MCP from the Claude directory, or add https://mcp.framer.com/mcp as a remote server in another client, then sign in to Framer and approve access, per the MCP guide. The agent works with your account's permissions.

Ask it to work on a branch, to rewrite only the variant's headline from your positioning notes, and to offer three options. The MCP page says changes stay in the project or branch until you or the agent publishes, and that an agent publishes only when asked, so say plainly that it must not publish. A person reads the three options, keeps one and merges the branch. Branching is a Pro feature, which this test needs anyway.

Step 4: Set the conversion step

Click Configure in the banner on the page. In the Analytics tab, click the + next to Steps and choose the event type, here the tracking ID from step 1. You can add several steps, but the guide uses the last one to calculate the conversion rate.

Step 5: Start, publish and deploy

Click Start, then publish. With staging switched on, deploy as well, or visitors are never assigned to a variant. Framer then splits traffic evenly, half each with two variants. To confirm it works before real traffic arrives, open the page on several devices, trigger the conversion yourself and watch it appear, as the troubleshooting article suggests.

Step 6: Read the result and keep your hands off

Framer calls a winner when a variant's probability of being ahead passes 90% and enough data has come in, and shows an estimate of the days left based on current traffic, per the A/B testing guide. Do not edit a running variant: the guide's advice is to stop the test, make the change and start a new one. With low traffic or similar variants, the result stays inconclusive, which is an answer too.

Step 7: Replace the control

When the test is conclusive, click Stop Test and pick the winning variant as the new control. Publish again so every visitor sees it, then delete the variants you no longer need.

What Framer's help pages warn about

• Published, not deployed. On a project with staging, publishing alone shows no results.
• A reverse proxy. Tests rely on the Server-Timing header, so a proxy that caches responses or rewrites headers breaks the assignment, per the troubleshooting article.
• Thin traffic. The guide recommends high-traffic pages; a quiet page may never reach a verdict.

Check the winner outside Framer

Before you click Start, write down the leads per week your CRM received from this page over the last two weeks. After the winner becomes the control, count the same thing for the next two weeks. That comparison is free, and it covers the buyers who converted after the first day. Framer crowns the winner in its own analytics on a one-day window, and the Convert add-on is billed on the events those same analytics count, so the counter that picks the variant is also the one that sets the invoice. Later, if you want that check run continuously against revenue rather than same-day clicks, a causal attribution tool like Causality Engine does it. Disclosure: I founded it.

FAQ

How many variants can one test have?

Up to five per page, each renamed so you can recognise it in the results, per the A/B testing guide.

Can I send the test's conversions to Google Analytics?

Yes. Framer fires custom events such as framer:click and framer:formsubmit that a script in the site's custom code can forward to Google Analytics or Segment, per the tracking guide.

How long should a test run?

There is no fixed duration. Framer shows the estimated days left from current traffic, and a test that stays level ends as inconclusive rather than with a forced winner.