Two things can make your tracker say "opened." A person reading your email. Or a piece of software loading it with nobody watching.
Same notification. Very different meaning. Here's the difference, in plain English.
What is a human open?
A human open is exactly what it sounds like: a real person, on a real device, looking at your email. It's the thing you actually want to know about, because it means your subject line worked and your message got attention.
Humans behave like humans. They open at normal hours. They read on phones and laptops, not on servers. And when an email matters to them, they come back to it. A prospect re-opening your pricing email the next morning is telling you something no other metric can.
What is a bot open?
A bot open is software loading your email's tracking pixel with no human involved. The pixel is just a tiny invisible image, and anything that fetches it registers as an "open." Machines fetch a lot of them.
Two sources do most of the damage:
Apple's pre-fetch. Apple Mail Privacy Protection loads tracking pixels on Apple's servers ahead of time, so every Apple Mail recipient looks like an opener even if they never read the email. Apple Mail is roughly 58 percent of email client share, per Litmus. Read that again: most inboxes auto-open everything you send them.
Security scanners. Corporate tools like Mimecast, Proofpoint, and Microsoft Defender open every link and image in incoming email, hunting for malware. Between 15 and 25 percent of B2B recipients sit behind one. Every email to those inboxes gets machine-opened on arrival.
There's also a middle case worth knowing: the [Gmail image proxy](/glossary/gmail-image-proxy). Gmail serves images through its own servers, so a Gmail open needs the full signal picture before you can call it human or bot. Counting all of them is wrong. Ignoring all of them is wrong too.
How do bots and humans behave differently?
Speed is the biggest tell. Bots fire within seconds of send, because scanners process email the moment it arrives. Humans take minutes or hours, and they open when they're awake.
Breadth is the second tell. A scanner hits many links at once, in a single pass, because checking everything is its whole job. A person clicks one thing, maybe two, with thinking time in between.
And return visits are the third. Bots touch an email once. Humans come back to re-read what interests them. Nobody's security scanner gets curious about your proposal overnight.
Why does this matter for your numbers?
Because most trackers add bots and humans into one number and call it your open rate. The result: raw open rates on B2B lists run 2-3x inflated. A dashboard showing 70 percent often means 25-35 percent real reads.
That gap changes decisions. You chase "engaged" prospects who never saw your email. You judge subject lines on data that's mostly Apple's servers. The metric feels precise and steers you wrong, which is worse than no metric at all.
Can you just block the bots?
No, and you wouldn't want to. These bots aren't attackers. Apple's pre-fetch exists to protect recipient privacy. Security scanners exist to keep malware out of company inboxes. They're doing useful work for the people you're emailing. Your open rate is just collateral damage.
You also can't opt out from your side. Apple decides how Apple Mail behaves, and a company's IT team decides what their scanner checks. No tracking pixel trick changes that. The bots will keep opening everything you send.
So the realistic goal isn't stopping bot opens. It's refusing to count them.
What's the fix?
Grade every open instead of counting every open. Outsolvi scores each one from Tier 1 (high-confidence human) down to Tier 5 (bot or scanner), based on request timing, network origin, and client signature. Anything under a 25 percent confidence floor is excluded from your counts. The grading rules are public on the [methodology page](/methodology), and the [confidence scoring](/features/confidence-scoring) page shows the tiers in action.
The grades hold up: the model agrees with human-rated classifications 95-98 percent of the time, and Tier-1 false positives run under 2 percent. When you want to check a specific open, the [Diagnose view](/features/diagnose-open-evidence) shows why it counted or got filtered. Every number opens up into its evidence.
The short version
Human open: real person, real device, normal hours, might come back. Bot open: instant, automated, no reading involved. Your raw open rate mixes them together, and the mix runs 2-3x too high on B2B lists.
If you want to see which trackers separate the two, start at the [comparison hub](/compare). Or try it on your own sends: Outsolvi is $7 a month billed yearly, with a 14-day free trial and no credit card. Two weeks in, you'll know how many of your openers were ever people.