Email sample
Edit the example or paste a message.
0 characters
An interactive answer to a stubborn question
Why advanced AI has not erased phishing
Detection is an economics and tradeoff problem. Filters must stop changing attacks without blocking legitimate mail, while attackers only need a tiny fraction of a massive, nearly free campaign to work.
1,000,000 sends
40 conversions
0.004% conversion can still clear a profit when delivery is cheap and automated.
Your text stays on this device. The analyzer never sends pasted email content or includes it in a network request.
Transparent heuristic analyzer
Inspect an email without a black box
Heuristics are clues, not proof. Review the sender, links, requested action, and context together.
Heuristic risk
Review recommended
0/100
Run the analyzer to see which visible signals contributed to the score.
Signal ledger
0 triggeredBase rates are brutal
When only a small share of all mail is malicious, an imperfect detector can flag many legitimate messages even with strong accuracy.
Attackers adapt
Every public defense creates feedback. Attackers rotate domains, wording, identities, infrastructure, and targets faster than static rules.
Cheap sends change the math
A campaign does not need to fool everyone. Automation makes tiny response rates economically meaningful.
False-positive tradeoff
Tune the filter, then watch who pays
Aggressive filtering catches more attacks, but it can quarantine more legitimate mail. No threshold is free.
PermissiveAggressive
0.1%10%
Confusion matrix
Simulated messagesActual class
Filter result
Phishing
Legitimate
Blocked
0True positive
0False positive
Allowed
0Missed attack
0True negative
Attacker economics
How a terrible response rate still pays
These are illustrative assumptions, not operational guidance. Change the inputs to see why volume matters.
Profitable at tiny scale
Estimated net$0
Delivered0
Conversions0
Gross value$0
Campaign cost$0
Gross value$0
Campaign cost$0
Layered filtering pipeline
AI is one layer, not a delete button
Modern email defense combines identity checks, reputation, content models, link analysis, behavior, policy, and human reports.
Stage 1 of 6
Five-question check
Can you reason past the obvious clues?
Strong security judgment is contextual: a polished message can be malicious, and an awkward one can be legitimate.
Question 1 of 50 correct
Choose the best answer.
The durable answer
Phishing survives because the system is asymmetric.
Defenders must make good decisions across billions of changing messages while preserving legitimate communication. Attackers can fail almost every time and still succeed. Better AI raises the cost of attack, but it also helps attackers vary copy and targeting. Identity, economics, user context, and recovery controls matter alongside content classification.