Your email open rate is lying to you (and has been since 2021)

If your open rate jumped some time in late 2021 and never came back down, nothing you did caused it. Apple did.

Mail Privacy Protection, announced at WWDC 2021, routes remote images through Apple’s servers and pre-loads them in the background. Your tracking pixel gets fetched for every Apple Mail recipient whether they read the message, glanced at it, or left it unread forever. As far as most email tools can tell, that is an open.

There is no protocol for "this person read your email". What exists is a workaround: put a 1×1 transparent image in the message body with a unique URL per recipient. When the mail app loads that image, the server records a hit. That hit is your "open".

Which means an open is really "something fetched an image". For a decade that was close enough to a human. It no longer is.

Apple Mail Privacy Protection inflates. It fetches for everyone, so a list heavy on iPhone users reports a higher open rate than an identical list on Android — with no difference in actual readership.

Image blocking deflates. Plenty of people and most corporate gateways refuse remote images by default. Those readers can open, read and act on your email and never register an open at all.

Plain text is invisible. A plain-text email has no image to load, so it cannot be open-tracked. Not poorly — not at all.

The distortions do not cancel out. They vary by list, by device mix and by industry, which is what makes open rate unusable as a comparison over time.

Clicks. A click requires a deliberate tap on a link, and no privacy proxy taps links on a reader’s behalf. It also works in plain-text email, where the pixel cannot.

That is why we treat click rate as the primary number in directinapp’s email tracking and label opens explicitly as an estimate. It is a less flattering headline. It is also the one you can act on.

You can separate a good share of the machine traffic if you look at the right signal. Gmail announces itself — its image proxy shows up plainly in the request. Apple does not: the Mail Privacy relay sends a user-agent indistinguishable from ordinary desktop Safari, so matching on the user-agent alone will miss every one of them.

The reliable signal for Apple is the network. Those fetches come from Apple’s own address space, which is stable and identifiable. That is how we classify them, and why our reported open rate does not climb simply because more of your list bought iPhones.

A tool that reports one blended open number, with no split, is either not looking or not telling you.

Lead with click rate. It is measured the same way for every recipient, so it is comparable across sends and across time.

Use opens for direction, not decisions. A send where opens collapse to near zero is worth investigating — something probably failed. A send where opens rise 4% is noise.

Never compare your open rate to a published benchmark from before 2021. Different measurement, different number.

And if you are running a small list — outreach, proposals, invoices — forget rates entirely and look at who. "Did this particular person click?" is a better question than any percentage across thirty people.

Open tracking is not useless. It is just far weaker than the industry pretends, and the pretending is the problem — people build forecasts and cut campaigns on a number that partly measures Apple’s infrastructure.

Say what it is, separate what you can, and put the reliable metric first. That is the whole trick.