Píxeles de seguimiento de apertura: qué miden realmente en 2026

The humble open tracking pixel—a 1x1 transparent GIF fetched from a server when an email client renders the message—has been a staple of email marketing for two decades. Yet in 2026, the signal is noisier than ever: Apple Mail Privacy Protection (MPP) pre-fetches images for all recipients, Google’s Gmail proxy caches pixels at the infrastructure level, and an increasing number of corporate Exchange servers strip remote content by default. The result is that a pixel “open” no longer means a human read your email—it often means a bot, a proxy, or a privacy relay did.

What a pixel actually fires on

When you embed an open tracking pixel in an email, the sequence of events is straightforward in theory: the recipient’s email client loads the HTML, encounters an <img> tag pointing to your tracking server, and makes an HTTP GET request. That request logs a timestamp, the recipient’s IP address (subject to anonymisation), the User-Agent string of the email client, and any query parameters you appended (e.g., campaign ID, recipient hash).

In practice, the “open” event fires under four distinct scenarios:

Human view: The recipient opens the email in a client that loads remote images by default (e.g., Gmail web, Outlook for iOS with images enabled, Thunderbird).
Apple MPP pre-fetch: Apple’s servers open every email in a background headless browser shortly after delivery, regardless of whether the recipient has opened it. This inflates open rates by 15-30% on iOS/macOS lists.
Gmail proxy cache: Google’s infrastructure fetches images for all Gmail accounts at the moment of delivery, caching them on Google’s CDN. The pixel request comes from a Google-owned IP, not the recipient’s.
Security scanner: Corporate gateways (Proofpoint, Mimecast, Microsoft Defender) often follow links and load images to scan for phishing payloads before the email reaches the inbox.

The critical distinction: only the first scenario represents a human reading your email. The other three are machine-generated noise. In 2026, depending on your list composition, 40-60% of recorded “opens” may be non-human.

The gap between open rate and read rate

Marketers have long used open rate as a proxy for engagement, but the proxy has decayed. A 45% open rate on a B2B list in 2026 could mean 15% genuine human reads and 30% Apple MPP pre-fetches, with the remainder split between Gmail proxies and security scanners. The raw number is misleading.

What a pixel cannot measure:

Time spent reading: A pixel fires once, at load. It does not track scroll depth, dwell time, or whether the recipient closed the email after one second.
Forwarding or printing: If a recipient forwards the email to a colleague who opens it in a different client, the pixel fires again—but you cannot distinguish the original recipient from the forwardee without additional tracking (e.g., a unique per-recipient link).
Offline reading: Desktop clients like Outlook for Windows (classic) and Apple Mail (offline mode) cache images locally. If the recipient reads the email while disconnected, the pixel never fires. When they reconnect, the client may or may not re-fetch the image.

The practical takeaway: open rate is a directional trend metric, not an absolute count. Compare week-over-week changes within the same list segment, but stop treating a single campaign’s open rate as a reliable engagement score.

Apple MPP and Google proxy: the two biggest distorters

Apple Mail Privacy Protection, launched in 2021, remains the single largest source of false opens in 2026. MPP routes all email images through Apple’s proxy servers, which pre-fetch the content in a randomised, isolated environment. The pixel fires for every email sent to an Apple Mail user who has MPP enabled—which, by default, is all iOS 15+ and macOS Monterey+ users unless they manually opt out. That is roughly 65% of Apple Mail users, and Apple Mail accounts for about 45% of all email client market share on mobile.

Google’s approach is subtler but equally distorting. Gmail does not pre-fetch images for every recipient at delivery; instead, it proxies all image requests through googleusercontent.com. The proxy fetches the image once and serves the cached version to the recipient. If the pixel request originates from a Google IP, you cannot know whether the recipient actually saw the email or whether the proxy simply refreshed its cache. Google also performs background link verification scans that can trigger pixel loads.

The combined effect: if your list is 50% Gmail and 30% Apple Mail (common for B2C), roughly 80% of your “opens” are at least partially machine-mediated. Only recipients on Outlook.com, Yahoo, ProtonMail, or custom domains with strict image-blocking policies produce relatively clean signals—and even those are subject to security scanners.

What you can infer from pixel data (if you filter correctly)

Despite the noise, open tracking pixels still provide useful signal when you apply server-side filtering. The key is to classify each pixel request by its characteristics before logging it as an open. Common filters include:

IP range blocks: Reject requests from known Apple MPP IP ranges (Apple publishes their ASNs) and Google proxy IPs. This cuts false opens by 40-50% but also misses legitimate opens from those providers.
User-Agent parsing: Apple MPP requests typically have a generic Mail User-Agent or no User-Agent at all. Gmail proxy requests carry a GoogleImageProxy User-Agent. Filtering these out leaves only requests from known email clients (Outlook, Thunderbird, Samsung Mail, etc.).
Timing heuristics: If a pixel fires within 2 seconds of delivery, it is almost certainly a pre-fetch or scanner. Human opens typically occur minutes to hours after delivery. Discarding sub-5-second opens removes a large portion of machine traffic.

After filtering, the remaining open rate—often called “verified opens” or “human opens”—is typically 50-70% lower than the raw rate. That filtered number is the one you should use for A/B testing subject lines, send-time optimisation, and list hygiene. A verified open rate below 10% on a warm list is a strong signal that your subject lines or sender reputation need work.

The one metric pixels still measure reliably

One thing open tracking does well, even in 2026, is relative comparison within a single send. If you send two variants of a subject line to the same list at the same time, the raw open rate difference between them remains valid—because the noise (MPP, Gmail proxy, scanners) applies equally to both variants. The absolute numbers are inflated, but the delta is real. This is why open tracking is still useful for A/B testing, even if you should not trust the absolute rate for reporting to stakeholders.

The regulatory landscape for tracking pixels has hardened significantly since GDPR took effect in 2018. In 2026, the following rules apply in most jurisdictions:

GDPR (EU/EEA): The pixel sets a cookie or logs a unique identifier tied to the recipient’s email address. This constitutes personal data processing. You need explicit consent before sending the tracking email, or a legitimate interest assessment that passes regulatory scrutiny (rarely successful for marketing emails).
ePrivacy Directive (EU): Requires consent for storing or accessing information on the user’s device. Since the pixel does not store anything locally, the directive’s application is debated—but most regulators treat the pixel as a form of tracking that requires opt-in.
CCPA/CPRA (California): The pixel does not trigger CCPA obligations unless you combine it with other data to build a profile. However, California’s 2024 amendments expanded the definition of “sale” to include sharing data for cross-context behavioural advertising. If you use pixel data to segment audiences for retargeting, you must provide a “Do Not Sell or Share” link.
Canada’s PIPEDA and Brazil’s LGPD: Both require meaningful consent for tracking technologies. Implied consent (e.g., “by subscribing you agree to tracking”) is no longer sufficient; you need a clear opt-in checkbox.

The practical impact: if you send to EU or UK recipients, you should include a plain-text notice in the email footer stating that the email contains a tracking pixel and provide a link to opt out of future tracking (or unsubscribe). Several ESPs now offer a “no tracking” flag per recipient that disables pixel injection.

How newmail.click handles this

At newmail.click, we build open tracking into our Pro tier with the understanding that the data is directional, not absolute. Our pixel server logs the raw request, then applies a server-side filter that discards requests from known Apple MPP and Google proxy IP ranges before counting an open. You see both the raw count and the filtered “verified” count in the dashboard. We also send Slack and email alerts when a verified open occurs for a specific contact—useful for sales sequences where a human opening your proposal is a trigger to follow up. For free-tier users, the signature builder includes no tracking by default; you can upgrade to Pro to enable click and open tracking, with UTM parameters auto-appended to links. We do not claim to solve every distortion—no vendor can—but we give you the tools to separate signal from noise without requiring a data science degree.

The open tracking pixel is not dead in 2026, but it has become a blunt instrument that requires careful calibration. Use it for A/B testing and trend analysis, not for absolute engagement scoring. Filter out the machine noise, respect privacy regulations, and treat the remaining signal as a rough indicator—not a fact.

Try the Pro trial — 7 days, no card