After we set up the Conversions API our lead numbers doubled. Are those real?

Why two senders exist at all

The browser pixel fires from the visitor's device and is subject to whatever the browser, the ad blocker and the operating system allow. The Conversions API fires from your server and is not. Running both is the intended design: the server send fills in what the browser lost, and Meta merges the overlap.

The merge is not automatic magic. It depends on you telling Meta that two arriving events are the same event. If you do not, Meta has no way to know that a server-side Lead and a browser-side Lead one second apart are the same person.

The three things that must line up

• Identical event name. A browser Lead and a server CompleteRegistration will never deduplicate, because Meta treats them as different events.
• Identical event ID. Generate one ID at the moment of the action and pass the same value to both sends. Generating one on the page and another on the server guarantees duplicates.
• Arrival inside the matching window. Server events that arrive long after the browser event may fall outside the window Meta uses to reconcile them. Batch uploads that run nightly are the usual offender.

How to verify in ten minutes

Submit a test lead yourself. In Events Manager, open the test events view or the event's detail and look at how many events were received and how many were deduplicated. Meta will show you both counts. A healthy setup shows a meaningful share of events arriving from both sources and being collapsed into one.

Then look at the event match quality indicator. Server events that carry hashed email, phone, name and location will match to people far more reliably than events that carry only an IP address. Low match quality does not create duplicates, but it does quietly degrade optimization, which is a slower and more expensive problem.

What it means for your reporting

Double counting inflates conversion volume and deflates cost per lead, which makes Meta look better than it is in exactly the place owners look. The correction is uncomfortable but necessary before you compare channels.

This is also an argument for judging channels on records that exist in your own systems rather than counts reported by the platform. When every Meta lead is matched to a real CRM record and a real job, duplicate pixel events become a tracking-hygiene issue rather than a reporting lie. That is the role of an integration layer and of reporting built on your own data.