Gmail, Outlook, Apple Mail: three providers, three different ways to break open rate

Most discussions of open-rate reliability focus on Apple Mail Privacy Protection, because it is the largest and most obvious distortion. But Apple is not the only actor. Gmail and Outlook have their own, smaller, opposing distortions — and the interaction between the three is what makes the aggregated number impossible to interpret.

This piece walks through each provider in turn: how their infrastructure treats tracking pixels, what the recorded open rate looks like compared to reality, and what you can realistically infer from each one in isolation.

Apple Mail: systematic inflation

Apple's MPP proxy fetches every remote image the moment a message arrives, and serves subsequent fetches from cache without triggering another hit on your server. The effect is total and deterministic: one open per delivered message, full stop.

The practical consequences:

• Reported open rate for Apple Mail recipients sits between 95% and 100% on any healthy list.
• Relative differences between Apple Mail segments are mostly noise — if one campaign opens at 98% and another at 96%, the gap is not meaningful.
• The one signal that survives is absence. If Apple Mail opens for a segment suddenly fall from 98% to 50%, something is wrong — delivery is failing, not attention.

Outlook: systematic suppression

Outlook — particularly Microsoft 365 enterprise tenants — does the opposite. The default image policy in many enterprise tenants is to block remote images, requiring the user to click "Download images" to load them. Until they click, the pixel never fires.

There is an additional layer: Microsoft Defender and Exchange Online Protection scan messages with what they call "Safe Attachments" and "Safe Links". The scanner fetches images as part of its rendering, but it does so from Azure IP ranges with a specific User-Agent, and some tracking systems filter those out as bots.

The net result for Outlook recipients:

• Reported open rate is typically 8–15% on a healthy list, even when engagement (clicks, replies) suggests higher.
• The number under-counts real reads by a factor of 3 to 5 in enterprise tenants.
• Consumer Outlook.com accounts are closer to Gmail behaviour, but still suppress opens relative to Apple.

Outlook on the desktop client

The classic Outlook desktop client (Win32) historically cached images in a preview pane, firing the pixel on preview rather than full open. This is still true in 2027 for users on Outlook 2019 and older. The newer Outlook-for-Windows behaves more like the web version.

Gmail: surface-dependent

Gmail is the most complicated of the three because its behaviour depends on where the user reads the message.

Gmail web

On the web interface, images are loaded through Google's image proxy (googleusercontent.com) when the message is opened by the user. Pre-fetch happens only for messages the user has already opened at least once — not for new messages sitting in the inbox.

Reported open rate on Gmail web is relatively honest: roughly 15–30% on a healthy list, correlating with real engagement.

Gmail mobile app

The Android and iOS Gmail apps pre-fetch images on message arrival for messages from the primary inbox, to enable offline reading. This was rolled out gradually from 2023 and now affects most users.

The result is that Gmail mobile opens are partially inflated — not as dramatically as Apple, but enough to make the number optimistic. Reported open rates on Gmail mobile lists are typically 10–20 percentage points above real engagement.

Gmail promotions tab

Messages in the Promotions tab are not pre-fetched by the mobile app. The signal there is cleaner: if you see an open from a Gmail Promotions recipient, a human actually opened it.

The aggregation problem

Now imagine a list that is 40% Apple Mail, 30% Gmail, 20% Outlook, and 10% other. Your ESP reports a single aggregated open rate.

What does that number mean? It is a weighted sum of:

• A near-100% Apple Mail number that means "delivered".
• A mixed Gmail number that partially reflects real opens.
• A suppressed Outlook number that under-reports real opens.
• A grab-bag of other clients, each with their own idiosyncratic behaviour.

There is no coherent quantity this weighted sum approximates. It is not "delivered" (because Outlook under-reports). It is not "read" (because Apple over-reports). It is not "engaged" (because it does not correlate with clicks or replies). It is a number with no referent.

What to measure instead, per provider

1. Apple Mail. Use inbox placement (where the message lands) and reply rate. Open rate is unusable.
2. Outlook. Use click-through rate, reply rate, and inbox placement. If you must use opens, accept that you are seeing 20–30% of the true engagement signal.
3. Gmail. Segment by tab (Primary / Promotions / Updates). Primary and Updates opens are semi-reliable; Promotions opens are trustworthy; pre-fetched mobile opens are inflated.

FAQ