How to read these numbers without fooling yourself
There are two benchmarks on this page and they are not the same thing. KnowBe4's Phish-prone Percentage is the share of employees who fall for an initial simulation before any training, which is why it is high (33.1% globally, from 67.7 million simulations). The Verizon DBIR median click rate is much lower because it pools clicks across all simulations, including heavily trained populations. If someone tells you "the average phishing click rate is 1.5%" and someone else says "it's a third of employees," they are both right and measuring different things.
Why click rate is a weak program metric
A single click does not prove ignorance. Attackers only need one person to succeed, while defenders need systematic resilience, so a low click rate can hide a brittle program. Click rate also says nothing about whether employees recognized and reported the attack, which is the behavior that actually contains an incident. Optimizing click rate alone can even backfire: punitive programs drive the number down on paper while people quietly stop reporting their mistakes.
What to measure instead
The stronger signal is the ratio of reported phishing to clicked phishing, a quality-of-defense measure, plus time to report. A workforce that clicks occasionally but reports fast and often is far safer than one with a pristine click rate and no reporting culture. Regulations like NIS2 now require proof that controls actually work, and a click-rate chart does not meet that bar. This is the difference between an activity metric and an effectiveness metric.
The behavior behind the clicks
The clicks themselves are downstream of behavior. CybSafe's Oh, Behave! 2025-2026 report found 65% of workers now use AI and 43% admit to sharing sensitive data through unapproved tools, which expands the attack surface faster than any single simulation captures. Benchmarks are a useful gut-check, but a mature program measures behavior and risk, not just who clicked a test email.