Email Open Rate in 2026: What It Still Tells You (and What It Lies About)

For most of email marketing's history, open rate was the metric. It was the first thing marketers looked at after a send, the primary KPI in performance reports, and the threshold that determined whether a subscriber was "active" or "lapsed." Open rate was not perfect — it required recipients to load images, which not everyone did — but it was directionally reliable enough to build entire marketing strategies around.

In September 2021, Apple Mail Privacy Protection launched and began pre-loading all email images — including tracking pixels — for Apple Mail recipients, regardless of whether they actually opened the email. In 2025, Gmail's Gemini AI added a second layer: auto-opening emails to generate summary cards. By 2026, open rate has become one of the most misleading metrics in email marketing — and programmes that still use it as a primary signal are making consequential strategic errors based on data that doesn't mean what they think it means.

This is not a guide about giving up on email analytics. It is a guide about understanding exactly what open rate now tells you, what it no longer tells you, and which metrics provide the reliable signals that open rate used to provide.

Open Rate: What It Was Supposed to Measure

Email open rate was always a proxy measurement. It measured whether a tracking pixel loaded — which was used as evidence that the recipient had opened the email and rendered the images. This proxy was imperfect from the beginning: recipients who had image loading disabled (common in corporate environments and with privacy-focused email clients) generated no open events even when they read the email carefully. Open rate undercounted genuine engagement systematically among privacy-focused audiences.

Despite this limitation, open rate was directionally useful for most senders for most of email marketing's history. The image-loading rate tracked reasonably well with actual human engagement, especially for consumer audiences using web-based email clients that loaded images by default. Open rates from different time periods and different sender-audience combinations were comparable enough that they could guide decisions about subject line quality, send timing, and list engagement levels.

The Apple MPP rollout in September 2021 broke this proxy relationship fundamentally — not by making open rate meaningless, but by changing what open rate measures. Post-MPP, open rate measures a combination of genuine human opens and machine-generated pixel loads, with no reliable way to distinguish between the two at the individual event level.

How Apple MPP Changed Open Rate Forever

Apple Mail Privacy Protection works by routing all email image requests through Apple's proxy servers. When an email is delivered to an Apple Mail account with MPP enabled, Apple's servers pre-fetch all images — including tracking pixels — at or shortly after delivery time. This generates a pixel load (an "open" event in the sender's analytics) regardless of whether the recipient has looked at the email at all.

The result: for every Apple Mail recipient with MPP enabled, the sender's analytics record an open at delivery time. The recipient might actually read the email five minutes later, or might delete it without reading it three days later — the analytics show an open either way. The open event is real in the technical sense (the pixel was loaded) but not in the human engagement sense (the recipient did not necessarily see the email).

The Scale of MPP Impact

As of 2025, Apple Mail holds approximately 50-57% of mobile email client market share globally. Among that user base, 40-56% have MPP enabled. This means that 20-32% of all email opens globally are now machine-generated Apple MPP events — not human opens.

The distribution is not uniform across audiences. US consumer email programmes with demographics that over-index on Apple devices (younger audiences, affluent segments, creative industries) may see 50-65% of opens as machine-generated. B2B programmes sending primarily to corporate Microsoft 365 environments (Windows + Outlook) have much lower Apple Mail penetration and correspondingly lower MPP impact. Any programme must assess its specific audience's Apple Mail adoption to understand the magnitude of MPP inflation in its own analytics.

Identifying MPP-Generated Opens

The most sophisticated ESPs (Klaviyo, Salesforce Marketing Cloud, Mailchimp, Brevo) provide MPP identification — tagging open events originating from Apple's proxy IP addresses as suspected MPP events, so they can be excluded from engagement calculations. If your ESP provides this feature, enable it immediately. The MPP-adjusted open rate (excluding identified Apple proxy opens) is a far more reliable engagement signal than the raw open rate.

If your ESP does not provide MPP identification, you can approximate it by examining the IP addresses in open event logs. Apple's proxy IP ranges are primarily Akamai CDN addresses — identifiable by reverse DNS lookup. Open events from these IP ranges are Apple MPP events; open events from other IP ranges are more likely genuine (though still subject to Gemini AI inflation, described next).

How Gmail Gemini AI Inflated Open Rate Further

In May 2025, Google rolled out Gemini AI summary cards as a default feature for Gmail users. Gemini generates a summary of email content to display in the inbox or at the top of opened emails — which requires loading the email content, including images, to generate the summary. This content loading fires tracking pixels, generating open events in the sender's analytics.

The Gmail Gemini AI open inflation layer operates differently from MPP: while MPP fires for all Apple Mail recipients with MPP enabled regardless of inbox placement, Gmail's Gemini AI apparently fires more selectively, based on whether the email is in a position where Gemini would be expected to generate a summary. The exact behaviour is not publicly documented by Google — but observable data from programmes monitoring their click-to-open ratios shows the pattern: open rates have increased for Gmail audiences since Gemini summary card rollout, while click rates have stayed flat or declined, causing a divergence in click-to-open ratio that indicates false opens are inflating the denominator.

The Diverging CTO Ratio: The Diagnostic Signal

The clearest indication that your programme's open rate is inflated by machine actions is a declining click-to-open (CTO) ratio over the 2021-2026 period. If your CTO ratio was 15% in 2020 and is now 8%, this is not evidence that your email content has become less compelling. It is evidence that the denominator of the CTO ratio (opens) has increased due to machine actions while the numerator (clicks, which require human action) has stayed roughly proportional to actual engaged audience size.

The CTO ratio decline is a diagnostic, not a problem to solve. The problem was the open rate inflation — the solution is to abandon CTO as a primary metric and adopt click-to-delivered rate instead (clicks divided by emails delivered, not by opens). This removes the inflated denominator entirely and provides a reliable engagement signal.

What Open Rate Still Tells You (Genuinely)

Open rate is not completely worthless — it still contains signal. Understanding what signal remains helps use it correctly as one of several inputs rather than discarding it entirely.

Directional trends within the same audience: If your open rate for Gmail recipients increases significantly over a short period without a change in sending volume or content, this is likely a genuine change (new subscribers from a high-engagement acquisition source, improved subject lines, better send timing). If it decreases significantly, something has changed in genuine engagement. The absolute number is unreliable, but large directional changes for the same audience over short time periods often contain real signal.

Comparison between audience segments (with caveats): Comparing open rate between two segments of your list that have the same ISP distribution (both primarily Gmail, or both primarily Outlook) can reveal real engagement differences between segments — because the machine-action inflation affects both segments roughly equally and cancels out in the comparison. Do not compare open rates between segments with different ISP distributions (one Gmail-heavy, one Outlook-heavy) — the different MPP exposure rates make the comparison meaningless.

Delivery confirmation for small audiences: For very small sends (under 1,000 recipients) or test sends, open rate still serves as a basic confirmation that the email was delivered and accessible. If zero opens are recorded for a 200-person test send, something went wrong at the delivery layer. The open event (even if machine-generated) confirms the email reached a point where its content was loaded.

Open Rate and Deliverability: A Complex Relationship

One of the most consequential misunderstandings in email marketing: the belief that open rate directly drives Gmail deliverability. The reasoning goes: "Gmail tracks opens as engagement signals → higher open rate → better engagement signals → better inbox placement." This reasoning is wrong in a specific and important way.

Gmail tracks genuine human engagement as a reputation signal — not tracking pixel loads. Gmail's own servers facilitate MPP-style proxy loading for its own Gemini AI operations, which means Gmail knows the difference between a Gemini-generated pixel load and a genuine recipient action. The engagement signals that feed Gmail's domain reputation model are based on genuine recipient behaviours that Google can observe in its own infrastructure: recipients starring emails, moving emails between folders, replying to emails, and similar actions — not third-party tracking pixel loads that any sender can use.

This means that artificially inflated open rates — whether from MPP or Gemini — do not artificially inflate your Gmail domain reputation. The reputation signals that matter are the ones Google controls and can validate as genuine. Open rate inflation in your ESP analytics is a measurement problem for your team, not a signal that misleads Gmail's reputation systems.

The corollary: genuine engagement still matters enormously for deliverability. Gmail is tracking genuine engagement — it just does it through signals it controls directly, not through your tracking pixels. Emails that recipients genuinely engage with (even if that engagement doesn't register as an "open" in your tracking system) generate the authentic engagement signals that build Gmail domain reputation.

The Replacement Metrics: What to Use Instead

The metric replacement framework for 2026 email analytics:

Suppression Without Reliable Open Rate Data

The suppression decision — when to suppress a subscriber who hasn't engaged — was traditionally made based on "no opens in 90 days." In the MPP era, this rule generates two types of errors:

Error type 1 — False negatives (suppressing genuinely engaged contacts): A subscriber using Apple Mail with MPP enabled may be reading every email in Gemini summary view, generating machine opens but no click events, and being genuinely engaged with the content. A "no clicks in 90 days" suppression rule would suppress this contact even though they are reading every email.

Error type 2 — False positives (keeping disengaged contacts active): A subscriber whose Apple Mail is generating MPP opens on every delivered email will show as "active" in open-based engagement scoring even if they deleted every email without reading it. A suppression rule based on "no opens in 90 days" will never trigger for this contact, even though they are completely disengaged.

The MPP-adapted suppression framework:

A/B Testing Subject Lines Without Open Rate

The most commonly asked question when marketers first encounter open rate unreliability: "If I can't use open rate for subject line A/B testing, what do I use?"

The answer depends on what subject line testing is actually trying to optimise for. If the goal is to find the subject line that gets the most people to open the email — and the email contains compelling content that drives clicks — then the downstream click rate captures the full impact of the subject line decision: a better subject generates more genuine opens, more genuine opens generate more clicks, higher click rate is the measurable outcome.

The practical challenge: click rate A/B tests require larger sample sizes to reach statistical significance than open rate tests, because click rates are typically 5-15x lower than open rates. With a 3% click rate, you need a significantly larger test audience than with a 30% open rate to detect a meaningful difference between variants. Calculate your required sample size at a statistical power calculator before running click-based subject line tests — and accept that some tests will require full-list deployment to a sequential cohort rather than simultaneous split testing if the audience is too small for reliable click-based evaluation.

For very small audiences where click rate testing is impractical, consider using open rate data only from non-Apple-Mail recipients (identified by email client tracking in the ESP) as a proxy — the non-MPP fraction of the audience provides a more reliable open signal than the full list, though this too is being reduced by Gmail Gemini AI inflation. The cleanest solution remains click rate, with appropriately sized test groups.

Reporting Email Performance to Stakeholders in 2026

The organisational challenge of open rate unreliability is not just analytical — it's communicative. Stakeholders who have received open rate reports for years have expectations calibrated to historical open rate benchmarks. "Our open rate dropped from 32% to 24%" reads as a problem even when the actual explanation is "we stopped counting Apple proxy pixels, which were 30% of our previously reported opens."

The transition to open-rate-independent reporting requires proactive stakeholder communication. The recommended approach: report both "total open rate" (for historical continuity) and "verified engagement rate" (click rate, reply rate, and conversion rate) as the primary performance metrics, with a clear explanation of why the primary metrics have shifted. Frame the change not as "open rate is broken" but as "we now have more accurate engagement data that better represents actual commercial outcomes."

The stakeholder conversation framing: "Open rate now includes a significant fraction of machine-generated events from Apple Mail and Gmail AI features that don't represent genuine subscriber engagement. Click rate, reply rate, and conversion rate are unaffected by these machine events and give us a more accurate picture of how our email is actually performing with real subscribers. We're transitioning our primary KPIs to these more reliable signals."