Where to find it
Open your page in the builder. At the top you'll see a Page Mode chip (it says Standard by default). Click it to switch views.
You get three options:
Standard — every visitor sees the same page. One page, one URL. This is the normal default.
Optimize · A/B Test — show two or three versions to random slices of your traffic and find the winner.
Personalize — show a specific version to a specific audience, based on UTM, language, country, or device.
Switching the view here doesn't change what your visitors see. You activate a mode with its own Start or Turn On button, and nothing goes live until you Publish.
Optimize (A/B Testing)
Reach for this when you're not sure which version performs better and want the data to decide.
Click the Page Mode chip and choose Optimize · A/B Test. The Variants panel opens on the left. Your current page is already there as Variant A, tagged CONTROL. That's your benchmark; every other variant gets measured against it.
Each variant has a ⋯ menu:
Rename — give the variant a clearer name
Duplicate — copy it as a starting point
Ask agent to edit — tell AI what to change
Set as control — make this the benchmark (the current control can't set itself)
Archive variant — remove it (you can't archive the control; promote another one first)
Adding a variant
Click + New Variant. You get two ways to build it:
Ask AI to generate — describe the angle you want to test and the agent drafts the variant for you. Quick suggestion chips give you a starting point, like "Lead with savings %, more urgent CTA" or "Shorter form, fewer fields above the fold".
Clone an existing variant — start from a copy of any variant and edit it by hand or with the agent.
Name it, create it, and repeat to add more.
Starting the test
When your variants are ready, click Set up test, then Publish. Your test is now live.
While it runs:
Traffic is split across the variants (each variant shows its split and conversion rate in the panel).
Each visitor keeps seeing the same variant on return visits, so their experience stays consistent.
Ending it: promote the winning variant. It takes 100% of traffic and becomes the page everyone sees once you publish.
You can have up to 3 variants per page.
Personalize
Use this when you already know you want certain visitors to see a tailored version, rather than testing to find out.
Click the Page Mode chip and choose Personalize. The Audiences panel opens on the left. Your current page is already there as Variant A, tagged FALLBACK. This is the Default, and it serves everyone who doesn't match a specific audience.
Adding an audience
Click + New Variant. This opens a two-step wizard:
Step 1 · Define audience rule — name the variant (say, "Google Ads"), then set a rule that decides who sees it. A rule is a field, an operator, and a value, like utm_source equals google. You can match on:
utm_source / utm_medium / utm_campaign — which ad or link the visitor came from
language
country
device (mobile / desktop)
referrer — the site they came from
Step 2 · Generate content — the wizard confirms the audience you just defined, then you build that version of the page in one of two ways:
Ask AI to generate — describe how this audience's page should differ from the Default and the agent drafts it. Suggestion chips help here too, like "Lead with the offer that matches this campaign".
Clone an existing variant — start from a copy of any variant and edit it later.
When your audiences are ready, click Set up, then Publish. Personalization is now live.
A visitor sees the first audience whose rules they match. If they match none, they see the Default. Order your audiences from most specific to most general.
Turning it off: you'll be asked which version becomes the single page for everyone (the Default is pre-selected). Your audiences are kept, so you can turn Personalization back on anytime and they'll all still be there.
You can have up to 5 variants per page: the Default plus 4 audiences. The panel shows your count, such as 1/5.
Seeing your results
Note: Per-variant analytics is coming soon. Once it's live, each mode will show a comparison table with visitors, conversions, conversion rate, and uplift for every variant, so you can tell which version is winning. For now, go ahead and set up and run your variants. The results view will arrive in a future update.
One rule to remember
Nothing you change is live until you click Publish. Switching modes, editing a variant, changing a split, adding an audience, starting or ending a test: none of it reaches real visitors until you publish. If a change isn't showing up, publish and refresh.
Common questions
Can I run an A/B test and Personalization at the same time? No. A page is in one mode at a time, so end one before turning on the other.
What happens to my original page? It's safe. In A/B testing it becomes the control; in Personalization it becomes the Default. Nothing is deleted.
Will visitors get confused seeing different pages? No. In A/B testing each visitor is locked to one variant. In Personalization each visitor only ever sees the one version that matches them.
Can I keep editing a variant after publishing? Yes. Edit it like any page, then publish again to push the update live.
How long should I run a test? Long enough to collect a meaningful number of conversions per variant. Wait until the difference is clear and stable rather than acting on a few early clicks. There's no fixed time.