What Mail Privacy Protection actually does
Starting with iOS 15 and macOS Monterey, Apple Mail routes every image in every message, tracking pixels included, through Apple's own proxy servers and downloads all of them the moment the message lands in the inbox. Not when the recipient opens it. Not when they glance at the notification. The moment it's delivered, whether they ever look at it or not.
Apple built Mail Privacy Protection specifically so a sender can't tell whether a real person opened an email using a pixel. It applies by default to anyone with "Protect Mail Activity" turned on in Apple Mail settings, which is most users, since it's an opt-out setting almost nobody actively considers.
What this actually breaks
- Open rate itself: inflated, and inflated unevenly, since it's Apple-only. The more iPhone- and Mac-heavy a list is, the further its "open rate" drifts from anything real.
- Send-time optimization: a tool deciding the "best time to send" from historical open timestamps is training on a pile of Apple pre-fetches that all land within seconds of delivery, not on when people actually read email.
- Re-engagement automations: a "hasn't opened in 90 days" trigger never fires for an Apple Mail user who hasn't looked at your emails in years, because Apple already faked an open for every single one.
- Subject line testing: if a meaningful share of "opens" happen identically regardless of what the subject line said, the test carries less signal than it looks like it does.
What to measure instead
Click rate hasn't been touched by any of this. A click still requires a human to see a link and decide to tap it, so it's the closest thing email still has to a real engagement signal.
- Unique clicks per send as the headline number in a report, not opens.
- Conversion events downstream of the click (a completed signup, a purchase), if there's tracking wired up that far, since those are even harder to fake than a click.
- If an ESP or analytics platform separates Apple MPP opens into their own bucket, exclude that segment before trending open rate over time, instead of blending it into one number that means less every quarter Apple's market share grows.
None of this is a template problem. It's an instrumentation problem, and the two get confused constantly: a real drop in engagement and a rising share of pre-fetched opens look identical on a dashboard that never separated them. Before assuming a redesign fixed, or broke, something, check whether the metric that moved was ever measuring what it looks like it measures.