If your team is celebrating a 70% open rate on cold outreach, you are celebrating machines.
Email open tracking has been quietly broken since September 2021. That is when Apple shipped Mail Privacy Protection in iOS 15, and the industry's go-to engagement metric became unreliable overnight. Most tools have not adjusted. Most Account Executives are still reading a number that is two to three times the real human-read rate, then making follow-up decisions on top of it.
This piece walks through the three forces that inflate open rates in 2026, the actual math on what your real number probably is, what the major tracking tools do (and do not do) about it, and the metric AEs should track instead.
What Apple Mail Privacy Protection actually does
Apple Mail Privacy Protection launched with iOS 15 in September 2021 and is now standard across iOS 15+, iPadOS 15+, and macOS Monterey and later. The first time a user opens the Mail app after upgrading, Apple prompts them to choose whether to enable MPP, and the prompt's recommended option is "Protect Mail activity." Most users tap it.
When MPP is on, two things happen behind the scenes. First, Apple pre-loads every piece of remote content in the email, including your tracking pixel, on Apple's relay servers before the user opens the email. Second, the IP address reported to the sender is Apple's relay, not the recipient's actual IP. The geolocation is meaningless. The open timestamp is meaningless. The user might never open the email at all.
Litmus puts Apple Mail at roughly 58% global email-client market share in early 2026. Even if only half of those users have MPP enabled (a conservative estimate), you are looking at roughly 30% of your list reporting machine opens by default. In B2B lists with high iPhone density, particularly tech, SaaS, and creative agencies, the share is higher.
Apple has [confirmed the behavior in its developer documentation](https://www.apple.com/legal/privacy/data/en/apple-mail/) and in iOS marketing materials. There is no technical workaround. MPP-distorted opens cannot be detected by counting pixel loads alone. They can only be detected by analyzing the pattern of the load: instant timestamp, Apple-attributable IP block, no follow-up click activity. Outsolvi flags these as Tier 4 with 18% confidence. Most tools count them as full opens.
Gmail's image proxy: the quiet second inflator
In December 2013, Gmail rolled out an image proxy that routes every embedded image through Google-controlled servers on the googleusercontent.com domain. The original justification was privacy. The side effect for tracking has been two-fold.
First, Google's image-fetching bot loads the tracking pixel before the human opens the email. This is fast (often within minutes of delivery) and happens for every Gmail recipient regardless of whether they ever read the message. Most tools see that fetch and record an open.
Second, because the image is cached on Google's CDN, the same pixel can be re-fetched as the user scrolls through their inbox or expands the email later. Each fetch can generate a separate open event if the tool does not deduplicate.
Outsolvi applies a 3-minute dedup window on Gmail-proxy traffic and a stricter pattern check on the User-Agent string. A pixel loaded by Google's image proxy within three minutes of the previous load is treated as one event, not two. Most tools do not apply this filter. Run the same campaign through Outsolvi and through Mailtrack and you will frequently see the open count differ by 30 to 60 percent on Gmail-heavy lists.
Corporate email scanners: the third (and stealthiest) source
This one is invisible until you go looking for it. Most enterprise email gets routed through a security gateway before delivery. Mimecast, Proofpoint, Microsoft Defender for Office 365, and Cisco Secure Email all pre-fetch every link and every image in every email to scan for malware, phishing, and known-bad infrastructure.
When a security gateway pre-fetches your tracking pixel, your tool records an open. When it pre-fetches every link, your tool records clicks. None of this is the human. The human has not even seen the email yet.
The signature is distinctive once you know to look for it. Scanners typically fire 3 to 8 pixel and link requests within a five-second window, from a single IP block, with a generic User-Agent. They also click every link, not just one, which is a behavior no real human exhibits. Outsolvi flags this as Tier 5 with 0% confidence and excludes it from open counts.
Most tools do not. Mailtrack's [public help documentation](https://help.mailsuite.com/) is explicit that every pixel load is counted. Yesware and Mixmax filter a partial scanner blacklist but do not surface confidence to the user, so an AE looking at a Yesware dashboard cannot distinguish a Mimecast-fetched open from a human one. HubSpot Sales Hub has more sophisticated filtering, but the pricing starts at $50 to $150 per seat per month, which is a different market than the $7 to $20 most AEs actually pay.
The math: what your real open rate probably is
Take a realistic B2B SaaS list. Assume:
- 50% of recipients are on Apple Mail with MPP enabled (conservative).
- 20% sit behind a corporate email gateway that pre-fetches images.
- 15% are on Gmail and trigger duplicate-fetch behavior.
Now suppose your tool reports a 70% open rate.
The MPP cohort contributes roughly 50% of recipients pre-loading at 100%, which is 50 percentage points of inflated opens. The scanner cohort contributes 20% pre-loading at 100%, which is 20 percentage points. The Gmail dedup distortion adds another 10 to 15 percentage points depending on how often the same pixel re-fetches.
Subtract those from the reported 70 and the actual human-read rate is closer to 25 to 35%.
On typical B2B lists with mixed Apple, Gmail, and corporate-scanner exposure, the gap between raw pixel-load count and Outsolvi's Tier 1 plus Tier 2 (high-confidence human) count is consistently 30 to 50 percentage points. The "we are getting 70% open rate" story is almost always a "we are getting 30% open rate plus 40% noise" story.
This matters because AEs make follow-up decisions on the wrong signal. They follow up the prospects who "opened three times" when those three opens were Mimecast's scanner running once per minute. They skip the prospect who looks dead but actually read the email on a desktop without MPP and is sitting on it. The cadence misses the buyer.
How most tools handle it (or do not)
Setting Outsolvi aside for a moment, here is how the major tools sit on filtering as of May 2026.
Mailtrack. Counts every pixel load as an open. No MPP detection, no scanner filtering, no confidence tier. The product was built for Gmail simplicity and has not evolved on the filtering side. Public docs confirm this. Most affordable in the category, and the inflated numbers are the price you pay. See [how Outsolvi compares to Mailtrack](/compare/mailtrack) for a side-by-side.
Yesware. Filters a partial list of known scanner User-Agents but does not surface confidence to the user. Open counts are presented as binary, opened or not. MPP-distorted opens count as full opens. Pricing $15 to $65 per user per month, which makes the lack of confidence scoring harder to defend. The [Outsolvi vs Yesware comparison](/compare/yesware) walks through the filtering gap in detail.
Mixmax. Gmail-only. Similar filtering posture to Yesware (some scanner exclusion, no confidence tier exposed). Better engagement-pattern UI than Yesware, but no fundamentally different filtering. $29 to $69 per user per month. The [Outsolvi vs Mixmax breakdown](/compare/mixmax) covers what is included at each tier.
HubSpot Sales Hub. Enterprise-grade scanner and MPP detection, the most accurate of the big-name tools by a clear margin. Filters are not exposed to the user, but the underlying open count is closer to reality. The catch is price: $50 to $150 per user per month for the tracking-included tiers. If you are paying enterprise prices, the filtering is largely solved. Most AEs are not paying enterprise prices.
Outsolvi. Confidence scoring on a 0-100 scale exposed as Tier 1 (80-100%, high-confidence human) through Tier 5 (0-20%, machine or bot). A 25% confidence floor means anything below that is excluded from open counts entirely. Three-minute Gmail-proxy dedup. Burst-pattern scanner detection. The point is not that the math is more complex (it is not), it is that the result is shown to the rep so they can act on it. $7 per user per month on annual billing, $12 per user per month on monthly.
What AEs should track instead
Open rate is the wrong North Star metric for outbound in 2026. There are cleaner signals, and the sales community has been writing about this for two years.
Reply rate weighted by reply latency. [Josh Braun](https://www.linkedin.com/in/josh-braun/), who writes on LinkedIn about modern outbound, returns to this point repeatedly: replies are signal, opens are noise. The variation worth adding is when the reply lands. A reply within four hours signals a buyer in the active evaluation window. A reply after seven days signals a cold lead who is being polite. The follow-up plan for each is different.
Click depth on high-intent pages. A click on your pricing page or your comparison-vs-Yesware page is worth ten opens. A click on your homepage is worth one. Outsolvi tags links by intent tier so reps can see which clicks predict closes.
Multi-touch engagement patterns. [Florin Tatulea](https://www.linkedin.com/in/floringtatulea/), who has written extensively about data-driven outbound, makes the case that single signals are unreliable and combinations are predictive. An open followed by a click followed by a forward, all within 48 hours, is a higher-confidence buying signal than any single metric in isolation. Outsolvi computes this as an engagement score from 0 to 100.
Forwards and replies-not-to-you. When a prospect forwards your email to a colleague, the colleague is now part of the buying committee. When they hit reply-all and bring in a procurement contact, the deal has entered the next stage. These events are visible to a tracking tool that watches the thread, not just the open.
The hot-lead threshold. Five opens plus two clicks within seven days, weighted by confidence tier. This is the trigger for Outsolvi's hot-lead detection and the trigger an AE should use for "drop everything and call this person now." [Sarah Brazier](https://www.linkedin.com/in/sarahmbrazier/) and [Will Aitken](https://www.linkedin.com/in/willaitken/) have both written about same-day follow-up on intent signals as the highest-leverage move in outbound. Outsolvi gives you a clean threshold to act on.
How to audit your current tracking today
Pull last quarter's data from your current tool. Look at the open count for any thread that did not result in a reply. If the open count is greater than five and the prospect did not click any link, the opens are almost certainly machines. Now look at the threads that did convert. The pattern is almost always 2 to 4 high-quality opens, 1 to 2 link clicks on the right pages, and a reply within 72 hours.
If your tool does not let you see the open pattern, just the count, you are flying with one instrument. That is the gap Outsolvi was built to close. [Try it free for 14 days](/pricing), no credit card, and run the same campaign through both tools for a week. The gap between reported and real opens is usually somewhere between 30 and 60 percent.
The metric that will move your number this quarter is not opens. It is reply rate inside the active evaluation window, and the tool that surfaces it cleanly is the one worth running.