Dark Pattern

The screens came first and the name came later. Put the cancel button in small grey type in a corner and the keep-my-plan button in the middle in large blue, and the number of people who cancel drops visibly. For years the person who brought that result to a meeting was told they had a good instinct for conversion. The number had gone up, so there was nothing to argue with.

The awkward part belonged to the people who found the screen distasteful. They could feel that something was off and had no word to point at it. Say "this doesn't sit right with me" in a review and the reply is "tell me exactly what the problem is." Nobody lied, no law was broken, and the button is right there on the screen. So the objection got filed as a matter of taste, and the same design shipped again the next quarter. The side objecting has a handful of angry reviews to cite; the side proposing has a table showing churn down by a few points. When only one side of a room holds numbers, the discussion ends the same way every time.

Naming the thing changed that balance. Around 2010 a user experience researcher in the United Kingdom gathered these designs under one label, and specific types soon split off with names of their own. Once the label existed, examples could be collected and laid side by side. When several hundred screens from unrelated companies landed in the same drawer, it became clear that this was not a matter of each company's taste but a set of moves that repeat. The sentence in the meeting changed too. Replace "this doesn't sit right with me" with "this is a roach motel" and the argument moves from feeling to precedent.

Regulation followed the vocabulary. Consumer authorities in Europe and the United States began looking at cancellation flows and consent screens, and structures where signing up takes two clicks on the web while cancelling requires a phone call started drawing orders and fines. Korean e-commerce rules cite the same examples whenever they are revised. At that point a dark pattern stops being an ethics debate among designers and becomes something consumer protection rules address directly. Consumer protection covers the rights that let a buyer undo a decision — cancellation windows, refunds, and the like — and a screen that deliberately narrows the path back out leaves those rights intact in form and empty in practice.

The design field splits into fourteen strands, and this is the only one named after something you are not supposed to build. Information architecture, interaction design, and UX writing all deal with how to make something. This strand deals with where the line sits between what can be built and what a team decides not to build even though it works. UX writing designs the words inside a product, and the same craft pointed the other way lands here immediately. Writing "No thanks, I'd rather waste my time" on a decline button takes exactly the skill that writing a clear error message takes.

Six terms sit under dark pattern, and all six belong to one strand: type. What separates them is which part of a person they work on. Group them three ways and the map holds still — bending the moment of choice, narrowing the way out, and taking something without the person noticing they gave it.

Three of them bend the moment of choice.

1. Confirmshaming — attaching an embarrassing sentence to the decline option. When "Start free trial" sits next to "No thanks, I'd rather waste my time," the two options do not weigh the same. Both buttons work, but pressing one of them means agreeing out loud to an insult about yourself. The cost is invisible in the interface and real in the moment.
2. Nagging — asking again after being told no. Show the notification permission prompt every time someone opens the app and by roughly the tenth pass some of them accept. The consent collected that way is not the result of changing one's mind; it is the result of clearing a window. So the opt-in rate climbs while the share of people who actually open a notification stays flat, and the team celebrates a number that measures fatigue.
3. Bait and switch — showing what looks like A and delivering B. The classic case is a close icon that starts an installation instead of closing anything. People act on familiar symbols without re-reading them, and this type steps on that habit deliberately. It also spends trust quickly, because the next time the same icon appears the person hesitates over every dialog the product shows.

Two of them narrow the way out.

4. Roach motel — easy to enter, hard to leave. The name comes from an insect trap that admits and does not release. Signing up takes thirty seconds on the web; cancelling means a phone call on a weekday afternoon, twenty minutes on hold, and three separate retention offers before anyone will process the request. Each step has a reasonable-sounding justification on its own, and stacked together they cause most people who set out to cancel to give up somewhere in the middle.
5. Forced continuity — a free trial rolling into a paid charge on its end date with no warning. Take the card up front, put the trial length and the first billing date somewhere inconspicuous, and a large share of users learn about the charge from a card statement. What makes this a pattern is not the charge itself, which the customer agreed to, but the decision not to send a message that could easily have been sent.

One of them takes what people did not mean to give.

6. Privacy zuckering — arranging settings confusingly so people hand over more personal data than they intended. Twelve consent items on one screen with only "Allow all" rendered as a large button, or a mix of switches where some mean off and some mean on so nobody can read the current state at a glance. Where the other five take money and time, this one takes something that does not come back. Deleting the account later does not retrieve data that has already moved.

The six are not separate tricks so much as layers that stack on a single flow. Take the card at trial signup, skip the reminder email, show three retention offers to anyone who tries to cancel, and label the final confirmation "Give up my benefits" — that is four of them wired end to end. This is why reviews work better as a walkthrough than as a checklist. Give a new hire an account, have them subscribe and then delete the account, and places that were invisible on paper show up. One line saying signup took one minute and cancellation took eleven convinces a room faster than six type names.

"You're proposing we move cancel to the third screen inside settings. That's a roach motel."

Agreement that the number will move, paired with a note that the method already has a name.

"Drop the reminder three days before the trial ends and conversion goes up. That money leaves again next month as refunds."

A warning that the short-term metric and the revenue that survives are moving in opposite directions.

"If the consent screen only gives Allow all a real button, it fails the next review."

The point is that this gets caught as a compliance issue before anyone has to argue ethics.

At Pagedeck the phrase came up while the team was reworking the flow from the free plan to a paid one. Ravi brought a projection showing double-digit conversion gains if the trial simply converted on its end date with no notice. Noa drew the flow screen by screen and asked one question: at what moment does the user find out they have been charged? The answer was the card statement.

Mia set one rule on the spot. If the moment a person pays and the moment they know they paid do not line up, the design does not ship. Grace reached the same conclusion from the finance side. Conversions won that way come back as refund requests and card disputes, and once the dispute rate passes a threshold the payment processor raises fees or drops the account. Pagedeck settled on emails three days before the trial ends and again on the day. Conversion landed below Ravi's projection. First-month refund requests fell by more than half, and a few people who had tried the product and left came back months later and paid. Someone who feels tricked at a payment screen does not return; someone who got a reminder and decided against it might.

The most common mistake is assuming bad intent is required. Nobody comes to work planning to deceive customers, so it is easy to conclude this does not apply here. The term is judged by outcome, not intent. Ask the person who built a five-step cancellation flow and every step has a reason. A confirmation was added to catch accidental cancellations, an exit survey gives the product team something to fix, and a retention discount is a genuine benefit to the customer. Five well-meant decisions combine into a maze that holds people who want to leave. So the test narrows to one question: how many more steps does it take to finish what the user came to do than to do the opposite? If signup is two steps and cancellation is seven, the gap is the answer. Nobody's motives need to be examined — put the two screens next to each other and it is visible.

The second mistake is treating the boundary with persuasion as a matter of opinion. Marketing moves people by definition, the argument goes, so where persuasion ends and manipulation begins is anyone's call. The blurry middle is real, but the questions that draw the line are not vague. Does what the user believes match what is true, and does choosing the other option cost roughly what choosing the recommended one costs? A banner saying three items remain is information when three items remain and a lie when the number is hardcoded. File this under taste and each decision tilts toward whichever side the quarter's numbers favor, and a few quarters later the product contains a maze nobody designed on purpose. Removing it takes longer than building it did, because every step arrives with the reason it was added and the metric that rose when it shipped.