Your tracker says a prospect opened your email six times and clicked three links. Sounds like a hot lead. It might be a security appliance in a data center.
That's a false positive: an open or click your dashboard reports that no human ever made. They're not rare glitches. On a typical B2B list, they're a huge share of everything your tracker records.
What is a false positive in email tracking?
It's a machine event counted as a human one. Your tracking pixel loaded, or your link got fetched, but the cause was software, not your prospect.
The tracker isn't lying, exactly. The pixel really did load. The problem is what the dashboard implies: that a person saw your email. Old-school trackers count every pixel load as an open because, years ago, that was mostly true. It isn't anymore. The [email tracking pixel glossary entry](/glossary/email-tracking-pixel) has the background on how pixel tracking works.
Who's opening your email that isn't a person?
Three main culprits, and you've got all three on your list right now.
Apple's servers. Apple Mail Privacy Protection pre-fetches tracking pixels on Apple's servers, 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. This is the biggest single source of fake opens, and the clearest: MPP pre-fetches are the most identifiable machine-open class there is, and they stay filtered.
Corporate security scanners. Tools like Mimecast, Proofpoint, and Microsoft Defender open every link and image in incoming email to check for malware. Around 15 to 25 percent of B2B recipients sit behind one. Scanners are the main source of fake clicks, which is worse than fake opens: a "clicked all three links" prospect who never saw your email.
Image proxies. Gmail serves images through its own proxy, which hides the direct device signal. Proxy opens aren't automatically fake, plenty are real reads, but they need judgment rather than blind counting.
How bad is the inflation?
Bad enough to change your decisions. Raw open rates on B2B lists typically run 2 to 3 times inflated. A reported 70 percent open rate is often 25 to 35 percent real reads.
Think about what that does downstream. You call the "engaged" prospect who was actually a Proofpoint appliance. You skip the quiet prospect who read your email twice but got lost in the noise. Your subject line tests are graded by machines. The whole feedback loop you're using to improve your outreach is polluted at the source.
Why can't trackers just filter this out?
The good ones can. It just takes more than a simple rule.
The naive fix is a blocklist: ignore opens from known scanner addresses, ignore Apple fetches, done. But providers change behavior constantly, machines get better at blending in, and a static rule written last year silently rots. Worse, aggressive filtering starts eating real opens, and now you've traded false positives for false negatives.
The tell is behavior. Real people open in patterns: normal hours, real devices, sometimes coming back to re-read. Machines fire instantly after send, often hitting many links at once. One event is ambiguous. The pattern around it usually isn't.
How does confidence scoring work?
Instead of a yes/no on each open, you grade it. Outsolvi scores every open from Tier 1, high-confidence human, to Tier 5, bot or scanner, based on request timing, network origin, and client signature. Anything below a 25 percent confidence floor is excluded from your counts.
The model agrees with human-rated classifications 95 to 98 percent of the time, Tier 1 false positives run under 2 percent, and it keeps learning as providers change behavior. The full grading approach is documented on the [methodology page](/methodology) and the [confidence scoring feature page](/features/confidence-scoring).
The part that matters most for trust: you can audit any call. The [Diagnose view](/features/diagnose-open-evidence) lets you click an open and see why it counted or was filtered. A number you can interrogate beats a number you have to believe.
How do you check your own numbers?
Start with a smell test. Is your open rate way above what reply rates suggest? Do opens land seconds after you send? Does one contact "click" every link in every email? Those are machine fingerprints.
Then compare tools honestly. When you [evaluate trackers](/compare), the question isn't which dashboard looks nicest. It's which one can explain any individual open on it.
If you want to see your real numbers, Outsolvi is $7 a month billed yearly, with a 14-day free trial and no credit card required. Run it alongside your current tracker for two weeks. The difference between the two dashboards is your false positive rate, made visible.