TL;DR — A phishing mail that works is not well written, it is well decided. Six building blocks decide whether it is opened: the display name (free text, no verification), the subject (must fit the recipient's day), exactly one link, a design that looks like the sender, a moment when the recipient is busy, and a landing page that does not ask for a password. Everything else — clever copy, urgency, threats — actively lowers the hit rate. In a measured simulation, the calmest subject line won; the loud ones were filtered before a human saw them. The template is below, anonymised and complete.
Starting position
I build these mails for a living — for clients' own domains, with their written approval, so that the people who protect the network find out before the people who sell the breach. What I learned over several campaigns is uncomfortable: the technical part is easy. DMARC, SPF, DKIM, a tracked link, an event endpoint — a day's work. The hard part is writing something a busy person opens without feeling manipulated afterwards.
So this article is not about tricking people. It is about the six decisions that make a mail credible, shown so that defenders can recognise them, and so that anyone commissioning a simulation knows what "realistic" actually means.
The pattern
Take any phishing mail that was actually clicked and strip the story away. What remains is a short list:
- A display name that matches someone the recipient already deals with. Free text. No certificate, no verification, no cost.
- A subject that is boring on purpose. "Neue Anfrage über Ihr Kontaktformular" beats "Dringend: Ihr Konto wird gesperrt" — the second one is a filter signature and a human alarm at the same time.
- One link. Not three, not a button and a footer link. One, and it goes where the text says it goes.
- A design the recipient has seen before. A mail that looks like the sender is not read, it is recognised.
- A moment. Tuesday 10:00, invoice week, the day before a trade fair. Context beats copy.
- A landing page that asks for nothing. The moment a password field appears, the test becomes an attack — and the detection rate jumps.
Everything that feels like pressure is a cost, not a gain.
The six building blocks, and what each one is worth
| Element | What it exploits | What a defender can check |
|---|---|---|
| Display name | Recognition, not identity | The address behind the name — always visible on tap, rarely looked at |
| Subject | Routine | Does it match an actual process of the sender? |
| Single link | Reading flow | Hover the link; the host after @ is the real one |
| Borrowed design | Recognition instead of review | Does the sender normally write HTML mails at all? |
| Timing | Attention | Nothing to check — this is where training helps |
| Landing page | Nothing, if it asks for nothing | A page with no form is a page that measures, not steals |
The last row is the one people misread. A simulation landing page is deliberately empty: no form, no input, no cookie, no third-party script. What is measured is the click, nothing else. If a mail you receive leads to a page that wants credentials, it is not a test — regardless of what the footer says.
Attack path
Note what is not in this chain: a compromised server, a leaked password, a zero-day. A phishing mail needs none of them. It needs a mailbox and patience.
The template, anonymised
This is the skeleton I use, reduced to the parts that matter. Every value that belongs to a client is replaced by example.tld. The link pattern is the classic user-info confusion: everything before the @ is a username, everything after it is the real host.
From: Muster Agentur <absender@ihre-domain.tld>
Reply-To: absender@ihre-domain.tld
Subject: Neue Anfrage über Ihr Kontaktformular
Preheader: 1 neue Nachricht · 16.11.2026
Guten Tag,
über das Formular auf Ihrer Website ist eine neue Anfrage eingegangen.
Die vollständige Nachricht mit Kontaktdaten und Projektumfang liegt
für Sie bereit:
https://kunde-beispiel.de@ihre-domain.tld/de/security-awareness/TOKEN?v=V1
└── was der Leser sieht ──┘└── wohin es wirklich geht ──────────┘
Rückfragen können Sie direkt auf diese E-Mail beantworten.
And the same mail as HTML, cut down to the structure — one table, one button, no images, no tracking pixels beyond the landing page:
<table role="presentation" width="100%" cellpadding="0" cellspacing="0">
<tr><td style="padding:20px 28px;border-bottom:1px solid #e4e4e7;">
<span style="font-size:13px;letter-spacing:.18em;">KUNDE</span>
<span style="float:right;font-size:11px;text-transform:uppercase;color:#71717a;">Website</span>
</td></tr>
<tr><td style="padding:30px 28px;">
<h1 style="margin:0 0 16px;font-size:20px;font-weight:600;">
Neue Anfrage über das Kontaktformular
</h1>
<p style="margin:0 0 22px;font-size:14px;line-height:1.7;">
über das Formular auf Ihrer Website ist eine neue Anfrage eingegangen.
</p>
<a href="https://kunde-beispiel.de@ihre-domain.tld/go/TOKEN"
style="display:inline-block;background:#18181b;color:#fff;
padding:12px 20px;text-decoration:none;font-size:14px;">Anfrage ansehen</a>
</td></tr>
</table>
Three details decide whether this lands in the inbox or in the quarantine:
- The display name carries the recognition, the address stays honest. A mail from your own verified domain passes SPF, DKIM and DMARC. Nothing is forged; the recipient simply does not look at the address.
- The link has no shortener. A shortened URL is a filter signal and a visible mismatch.
- No images at all where possible. An HTML mail without remote images has nothing to block, nothing to warn about, and weighs 2 KB.
Why the design matters more than the copy
The most effective variant in the last campaign was not the cleverest one — it was the one that looked like a real agency mail because it had been measured from a real agency site: ground colour, type scale, headline weight, letter-spacing, the hard-edged button, the tiny uppercase labels. Same content, different recognition.
That has a defensive consequence: "we changed the password" does not help against a mail that never asked for one. What helps is the boring question — did I expect this mail from this sender, in this shape, today?
Detection
- Address, not name. Tap the sender. The display name is decoration.
- Hover the link. Everything up to the
@is a username and can be anything. - Shape of the mail. Does this sender normally send HTML with a single button?
- The missing form. A page that explains itself instead of asking for credentials is a measurement — or a decoy.
- Your own DMARC reports. If nobody can send as your domain, half of these mails never arrive.
Defence
What actually lowers the number, in this order:
p=reject, plus DKIM, plusruareporting. Not quarantine — reject.- One training a year is not training. The two-minute rule beats the annual slide: address, link, expectation.
- A reporting button that a human answers. People who report and hear nothing stop reporting.
- Simulations measured honestly — transport and clicks, never "how many employees failed", and never without the recipient's knowledge of the programme.
Once an attacker gains access, they can perform the following actions:
Conclusion
A phishing mail is a template plus six decisions, and five of them cost nothing. That is why it works and why "more technology" alone does not fix it: the technology can only remove the sender — the display name, the subject, the timing and the recognition stay human problems.
The useful question for a defender is therefore not "would I fall for this?" but "would I have noticed that this mail has no reason to exist today?"
Note on method: all examples in this article are anonymised. Domains are example.tld, names are roles, and no real campaign, client or mailbox appears. The measurements come from simulations run on domains whose owners commissioned them.