An open rate no longer tells you how many people read your email. Since Apple's Mail Privacy Protection arrived in late 2021, Apple's servers download the tracking pixel for anyone who has it switched on, whether they look at the message or not. What the number can still do is warn you when something breaks, as long as you watch it against itself and on the slice of your list Apple doesn't touch. What it can no longer do is pick a subject line, choose a send time or decide who counts as engaged.
How much of the number is Apple
Litmus counts the email clients behind more than a billion and a half opens a month. In its September 2026 report, Apple accounted for 76.56% of them, a figure that includes Mail Privacy opens and Apple's own apps. Not every one of those is a machine fetch, but Litmus itself treats privacy-protected opens as unreliable, and no tool on the sending side can tell a real Apple open from a prefetched one.
So on a typical consumer list, most of what your platform calls an open is a record that an email reached an Apple inbox. Those opens also arrive in a burst minutes after the send, which is why open rates jumped across the industry in 2021 and why a 45% open rate today says less than a 20% one did in 2020.
Klaviyo tags every open with a property called Apple Privacy Open, true or false, on every Opened Email event since 20 November 2021. That one field is what makes the number usable again.
What the number still tells you
That something changed. Your own open rate, on the same kind of send, week after week, is still a working alarm. Apple's share of your list doesn't swing overnight, so a sharp fall usually means something real: mail sliding into spam at one provider, a broken sending domain, a list import full of dead addresses. The Gmail and Yahoo sender rules are the usual suspect when the fall is sudden.
Where the drop is. Filter the opens to Apple Privacy Open equals false and you are left with Gmail, Outlook, Yahoo and the rest, where an open still mostly means a human loaded the images. That rate is lower and noisier, and it is the one that moves when inbox placement does. If it falls and the Apple-inflated headline holds steady, the headline is hiding a deliverability problem at a provider that matters.
What it can't tell you anymore
Who is engaged. The classic engaged segment, "opened in the last 30 days", now includes every Apple user who received an email, including the ones who never read anything you send. Swap the condition to clicked, or opened with Apple Privacy Open set to false, or placed an order, and widen the window to make up for the smaller signal. How wide depends on how often you send, which we worked through in your engaged segment is probably too small.
Which subject line won. This is where the damage is worst, and most of the advice still skips it. A subject-line test on opens measures which variant Apple's servers fetched, and they fetch both.
Strip the Apple opens out and the test gets honest but small. We ran the arithmetic on a real account with a list of about 7,000. Split in two, each variant went to 3,712 people, and the smallest difference a two-proportion z-test could separate from noise was a 35% relative lift. That was the floor with every recipient counted. Keep only the non-Apple opens, roughly a quarter of the total on Litmus's count, and the floor roughly doubles, because the detectable difference scales with the square root of the sample. No subject line beats another by 70%. On a list that size, a subject-line test on opens can't produce a winner, it can only produce a number that looks like one.
The rule we took from that account: on a small list, stop testing things that move opens and test things that move orders, like the offer, the angle or the segment, and judge them on clicks and revenue.
When to send. Send-time optimisation built on opens learns when Apple's proxy fetches images, which is shortly after delivery, whatever time you pick. Judge timing on clicks and orders, or leave it alone.
What to read instead
Clicks still need a person. Unique click rate per delivered email is the closest thing left to the open rate's old job, and it isn't inflated by Apple. Click-to-open rate is the exception, since its denominator is the inflated open count, so it now drifts down as Apple's share drifts up.
For anything about money, read revenue per recipient. It is the one email metric that goes to zero when an email stops earning, and it has its own traps, which we covered in revenue per recipient, the one flow metric worth arguing about. One of them comes straight from this post: Klaviyo's default attribution counts opens, so a privacy open can claim an order the customer was going to place anyway.
Are open rates still worth reporting?
Yes, as a trend line on your own sends and split by Apple Privacy Open. As a headline figure for a client or a board, no. It rises with Apple's market share and says nothing about whether anyone read the email.
Should I remove Apple Mail users from my sends?
No. Many of them read and buy. Exclude people who show no sign of life beyond privacy opens for a long stretch, meaning no click and no order, and keep everyone who does either.
Why did my open rate jump in late 2021?
That was the rollout. Apple's privacy feature shipped with iOS 15 in September 2021, and Mail asked every user to choose on first launch after the update. Opens rose within weeks with no change in how many people were reading.
