Gmail openings crashing: What's happening?

The scenario has come up from time to time in our client discussions in recent weeks: campaigns launch normally, content hasn't changed, neither has the list, and yet the opening rate The open rate on Gmail is plummeting from one campaign to the next. It’s not just a small drop—it’s more like a 10%. It’s actually halving, or even more.

When Gmail accounts for a significant portion of the user base—often 40% to 60% in the B2C sector in France—the impact on overall metrics can be immediate. The dashboard gives the impression of a sharp decline in performance, even though nothing has actually changed in terms of content, targeting, or reputation.

This is a highly variable behavior, not all of our customers are affected. Far from it.

What is observed in the data

On a very recent client case, we looked at three consecutive campaigns, a few days apart:

• Campaign 1 (April 27) Gmail open rate “normal”, in line with historical data.
• Campaign 2 (May 13) - the Gmail clicks are higher than on the previous campaign, but the openings are divided by more than two.
• Campaign 3 (May 18) — a few days later, the Gmail openings climbed back up to almost the pre-level, without us having changed anything.

Two things stand out:

1. The clicks, they have remained stable (and even increased). So emails are arriving, being read, and generating engagement.
2. The variation focuses on Gmail, not on other webmails.

In other words: it's not the campaign that lost performance, it's the measurement who started lying a little more than usual.

Why is it (almost) never a deliverability problem?

When an open rate drops, the first reaction is a healthy one: we wonder if we're not entering spam territory, especially if it's limited to one destination. But a few signals can quickly rule out this hypothesis:

• The clicks remain stable They're making progress. An email that gets clicked is an email that gets seen.
• The complaint rate and the bounce rate do not evolve.
• The Google Postmaster Tools do not show a decline in reputation.

In this context, the diagnosis is clear: it's not a deliverability issue, it's an issue of tracking of the opening, as confirmed by Mailgun «The most likely explanation is a measurement shift, not a deliverability problem.

What's (probably) happening on Google's side

For several weeks, several market players have been reporting the same phenomenon. Hypotheses are converging around Google infrastructure changes that affect how the tracking pixel is called (or not) when emails are displayed.

It should not be forgotten that openness, as it is measured today, relies on:

• a 1x1 pixel image loaded at email opening,
• or, in the case of Gmail, on a image proxy who downloads and caches images on the Google side.

This system is fragile by design: it depends on the behavior of the mail client, the proxy, preloading, and privacy protections (Apple's AMPP is the best-known example).

Gmail prefetching in practice

Technically, Gmail doesn't just serve images via its proxy: it also preload a portion when the user has an active Gmail session (web or mobile), even before they actually open the email (see this Sparkpost article for more technical detailsThe tracking pixel is therefore sent at that moment and is counted as an “open” by the ESP, even though no one has read the message yet.

Second layer: it would seem that Gmail has modified (at least temporarily) the way this prefetch works, by becoming more selective about the emails it decides to pre-load. Result: fewer pre-loaded images? fewer pixels triggered? fewer opens recorded, without any user behavior having changed.

This isn't an isolated rumor; several deliverability experts are documenting it:

How many “opening” points are in play?

The orders of magnitude available in the literature converge:

• +1 to +2 points inflated open rates due to prefetching according to SparkPost / Litmus (historical estimate across all traffic).
• +1 to +6 points according to MailSoar, depending on Gmail's share of the base and recipient behavior.
• Impact side of the recent decline : Validity measures a drop of approximately one third of the activity Gmail image loading November 2025, and some senders are seeing A drop in the Gmail open rate of 30% over a quarter or more.

In other words: when your database is at ~50% Gmail, losing a large portion of the prefetch can Mechanically erase 2 to 5 points of overall opening rate, with no loss of real engagement. That's exactly what we see on the ground, opens down, clicks stable, conversions stable.

The real subject: openness is no longer a reliable metric

This Gmail incident is not an isolated accident. It's a nth illustration on the fragility of the open rate:

• Apple Mail Privacy Protection (AMPP) Preload images without intent to read? Artificially inflated openings.
• Gmail caches images and can pre-load them even without opening the recipient.
• Anti-tracking protections businesses and browsers continue to strengthen with mass bot openings.

Result: compare its opening rate to that of 6 months ago, or to that of another Sectoral email benchmark, This is like comparing two thermometers that don't measure the same thing.

Why doesn't it affect everyone?

To prepare this article, I interviewed several deliverability experts of campaign management solutions and other ESPs (mainly French), to ask them: “Do you know if, in the calculation of opens, you «clean» Gmail prefetch opens?”

The answer is overwhelmingly yes. But what emerges is that this cleaning of statistics is very uneven. Some players clean without saying anything. Other players display raw stats and cleaned stats. Others actors don't clean anything at all (and therefore have very, very skewed statistics).

Short testimonial from Florent Destors, Deliverability Manager, Selligent/Zeta

“Regarding the changes in Gmail opens, we noticed that not all of our clients were impacted in the same way. You have the changes to prefetching, but you also have the introduction of the sorting option in the promotions tab (by date/relevance). From what I've observed, senders who target a broader audience are most likely to see significant variations.”

Another deliverability manager also confided in me that it caused them problems when some clients compared performance across platforms. Cleaning up these interactions was to their disadvantage, when it allows them to be a bit closer to reality.

So, the reason it doesn't affect everyone the same way lies halfway between your engagement levels (which impact Google's prefetch rate) and the cleaning (or not) of your statistics by your email sending tool.

What to pilot instead

The goal is not to abandon measurement, but to move the center of gravity towards more stable signals and closer to the business value:

• The click — less manipulable by ISPs, closer to the real intention. It is today the most robust engagement metric (at least in B2C)
• Single click per segment / per ISP — to detect targeted issues (e.g., “my Microsoft clicks are down while Gmail is stable, the problem lies elsewhere”).
• Downstream conversion — Qualified visits, sign-ups, purchases, completions. That’s what we’re really looking for as campaign results.
• Hard deliverability metrics“ — complaints, bounces, reputation, inbox placement (via seedlists).
• The unsatisfaction rate and inactivity — to manage the pressure and quality of the database.