The default arm is the real page
Opting out reliably falls back to Default for audiences, but a live A/B split still buckets an opted-out visitor into a real variant.
Default isn't a placeholder arm — it's resolved and served exactly like any other, and it's what a visitor gets whenever nothing stronger claims them: no audience match, no live experiment, no locked winner. Knowing precisely when that's true matters more than it sounds, because "nothing personalized this visitor" is a claim you may need to stand behind.
Opting out reaches audiences, not a running split
A visitor who opts out has their audience match forced to none at the point the audience engine resolves it, regardless of what a rule or a score would otherwise have matched — so every audience-driven piece of the page, section-level and module-level alike, falls through to Default.
No crawler gets special treatment
There is no crawler or bot detection anywhere in the audience or experiment code — nothing branches on user agent. A crawler is treated exactly like any other anonymous visitor: it lands on Default only because it carries no cookies or signals to match an audience against, the same reason a first-time visitor with nothing set yet would.
A visitor with no JavaScript gets the same real assignment
Assignment — audience match, experiment bucket, or Default — is fully resolved on the server before the page ever reaches the browser. A visitor who never runs client-side JavaScript isn't routed to some degraded fallback; they get exactly the same resolved arm a JavaScript-enabled visitor with the same signals would.
Pausing an experiment reverts the container to Default
When a page is rendered, the only experiments loaded at all are the ones currently running. An experiment you have left in draft, or have paused, is simply not among them — so no assignment is made for its container, and the container falls all the way through to the Default arm.
It is easy to expect a paused experiment to hold everyone on its control arm, since that is the arm the test was measured against. It doesn't. Control and Default are two different things: control is whichever arm of the experiment you marked as the baseline, and Default is the container's own unpersonalized arm. They are the same content only if you set them up that way. So if you pause a test to "put the page back to normal," check what the Default arm actually says — that, not the control copy, is what visitors will see.
What actually lands someone on Default
- No audience matched, or the visitor opted out.
- The container's experiment isn't running — a draft or paused experiment contributes no assignment at all.
- No winner has been locked for it.
See Which version of a page a visitor gets for how those checks are ordered against each other.