Somewhere in your product, a designer once made a button hard to find on purpose. Maybe the "cancel" link, maybe the "no thanks." It probably nudged a metric in the right direction for a quarter. It also taught your customers that you'd trade their trust for a conversion, and that lesson sticks longer than the quarter does.

The quick version

  • A dark pattern (now often called a "deceptive pattern") is an interface deliberately designed to trick people into choices they wouldn't otherwise make, hidden costs, forced sign-ups, cancel flows built like a maze.
  • The test isn't "did it work." It's whose interest the design serves: a nudge helps the user decide; a dark pattern overrides their decision.
  • This is no longer only an ethics debate. The EU's Digital Services Act and the US FTC now treat dark patterns as a legal problem, with live (if still uneven) enforcement.
  • You don't avoid them with a values poster. You avoid them by changing what you measure and by giving someone permission to say "that's a dark pattern" in the room.

How to tell a nudge from a trick

The term is younger than you'd think. UX designer Harry Brignull coined "dark patterns" on 28 July 2010, registering a site to name and shame interfaces "carefully crafted to trick users into doing things." His deceptive.design catalogue (he now prefers "deceptive patterns," partly to retire the loaded word "dark") gives the tricks plain names: confirmshaming ("No thanks, I like paying full price"), roach motel (easy to get in, hard to get out), sneak into basket, disguised ads. Naming them is half the work. Once a tactic has a name, a team can no longer pretend it's just "optimisation."

So put the vocabulary on the wall. When everyone on a product team can say "that's a roach motel" or "that's confirmshaming," the conversation shifts from a vague feeling that something's off to a specific, nameable thing you can choose not to ship. Borrow Brignull's taxonomy wholesale; you don't need to invent your own.

It's measurably common, and that's the honest problem

The strongest evidence that this isn't a fringe worry comes from a 2019 study by Arunesh Mathur and colleagues at Princeton and Chicago, "Dark Patterns at Scale" (Proceedings of the ACM on Human-Computer Interaction, 2019). They crawled roughly 11,000 shopping websites and found 1,818 instances of dark patterns across about 1,250 sites, sorted into 15 types and seven categories, and noted that the more popular the site, the more likely it was to use them.

For a working checklist, though, the cleaner frame is the five-strategy taxonomy from an earlier paper, Colin Gray and colleagues' "The Dark (Patterns) Side of UX Design" (CHI 2018). Their five families are worth memorising: nagging (repeated interruptions), obstruction (making a task needlessly hard), sneaking (hiding or delaying information), interface interference (visual tricks that privilege one option), and forced action (demanding something irrelevant to get what you came for).

The honest limitation: that study counts the presence of patterns, not their effect on any individual. It doesn't prove a given timer cost a given shopper money, and "dark pattern" sits on a spectrum, one person's helpful default is another's sneaky pre-tick. The research tells you the tactics are everywhere; it doesn't excuse you from judgement about your specific case.

So run those five families as a checklist against your own funnel. Walk your sign-up, checkout and cancel flows and ask, at each screen: is this nagging, obstructing, sneaking, interfering, or forcing? You will almost certainly find at least one. Most teams do.

flowchart LR
    A(["A design choice
boosts a metric"]) --> B{"Whose interest
does it serve?"}
    B -->|"Helps the user
decide for themselves"| C(["Ethical nudge
(default, reminder)"])
    B -->|"Overrides the user's
own decision"| D(["Dark pattern
(trick, trap, hide)"])
    C --> E(["Trust compounds"])
    D --> F(["Short-term lift,
long-term distrust + risk"])
The same metric lift can come from a nudge or a trick. The fork is whose interest it serves. Leaders Loop

Where the line actually sits: nudge vs. trick

The useful distinction comes from behavioural economics. Richard Thaler and Cass Sunstein's Nudge (2008) defines a legitimate nudge as one that steers people while preserving freedom of choice, and Thaler's own rule of thumb, repeated often, is "nudge for good." A default that enrols you in a pension you can leave with one click is a nudge. A pre-ticked box that signs you up for marketing you have to hunt to escape is a dark pattern. Same mechanism (a default); opposite ethics. The difference is whether the design is on the user's side or against it.

This is also where design ethics stops being abstract and becomes a measurable property. The cleanest test in practice: would the design survive being explained to the user out loud? "We pre-selected the expensive plan because most people don't read this far" is a confession. "We defaulted you to the plan most people in your situation choose, and switching is one tap" is a defensible nudge. If you'd be embarrassed to narrate it, it's a dark pattern.

A nudge helps you do what you already wanted. A dark pattern makes you do what the company wanted.

The practical version: write the one-sentence honest narration for any design that touches money, consent or cancellation. If the sentence sounds like a trick when read aloud to a customer, you've found one. Change the design, not the wording of the sentence.

The law caught up, this is now a risk, not just a value

For years, "don't be deceptive" was a matter of conscience. It isn't only that now. The EU's Digital Services Act, fully applicable across the EU from 17 February 2024, is the first EU law to name dark patterns directly: Article 25 prohibits online platforms from designing interfaces that "deceive or manipulate" users or otherwise impair their ability to make free, informed choices. In the US, the FTC's September 2022 staff report "Bringing Dark Patterns to Light" put companies on notice and catalogued four recurring tactics it treats as unlawful: disguised ads, hard-to-cancel subscriptions, buried fees and tricked-out data sharing.

The honest caveat: enforcement is uneven and the definitions are still being argued in court and in scholarship, so "is this technically illegal in my jurisdiction?" is a question for a qualified lawyer, not a UX checklist. But the direction is one-way. Designing a deliberate roach motel today is taking on regulatory risk you didn't have five years ago.

So treat your cancellation and consent flows as legal surface area, not just conversion surface area. The dark pattern drawing the most regulatory attention right now is making it harder to cancel than to subscribe; it's named in both the DSA and the FTC material. If your "subscribe" is one click and your "cancel" is a phone call, fix that first.

A worked example

Take a subscription media app whose growth team is under pressure on trial-to-paid conversion. (Figures below are illustrative.) Someone proposes three changes: a pre-ticked "auto-renew at full price" box, a countdown timer on the trial-upgrade screen reading "offer ends in 09:58," and a cancellation flow that routes through four screens of "are you sure?" and a retention offer before the actual button.

Each one would probably lift conversion next month. Each one is also a textbook dark pattern: the pre-tick is sneaking and interface interference, the fake timer is sneaking (the offer doesn't really expire), and the four-screen cancel is obstruction, a roach motel. Run them through the honest-narration test and they all fail: "we hid the real price behind a default," "we invented a deadline," "we made leaving exhausting."

Now the ethical version of the same goals. Replace the pre-tick with a clear, unticked choice and a one-line explainer of what each plan costs. Drop the fake timer; if there's a genuine launch discount with a real end date, say so truthfully. Make cancellation one screen and one button, and put a single, honest "before you go, here's what you'll lose" message after they confirm, not as a barricade before it. Say trial-to-paid dips a few points in the first month but refund requests, chargebacks and one-star "I couldn't cancel" reviews fall, and ninety-day retention of the people who do convert rises because they chose it on purpose. The honest funnel converts slightly fewer people into customers who actually stay.

flowchart TD
    A(["Goal: lift trial-to-paid"]) --> B{"Which lever?"}
    B -->|"Pre-tick + fake timer
+ 4-screen cancel"| C(["Metric up this month"])
    C --> D(["Chargebacks, 1-star
reviews, churn, legal risk"])
    B -->|"Clear price, honest
deadline, 1-tap cancel"| E(["Metric flat / slightly down"])
    E --> F(["Higher 90-day retention,
fewer refunds, trust"])
Illustrative: two routes to the same goal, with opposite second-order effects. Leaders Loop

The reframe that makes this stick: a dark pattern doesn't create demand, it borrows it from the future at a high interest rate. The same diverge-then-converge honesty that underpins design thinking & the double diamond applies here, solve the user's actual problem rather than trapping them inside yours.

Frequently asked questions

Isn't every persuasive design a bit manipulative?

No, and that slippery-slope framing is how teams talk themselves into bad choices. Persuasion presents a real benefit clearly and lets you decline easily; manipulation hides information, fakes urgency, or punishes the choice the company doesn't want. A good gut check is reversibility and transparency: can the user see what's happening and undo it in one step? A nudge passes; a dark pattern fails.

What's the difference between a "dark pattern" and a "deceptive pattern"?

They're the same thing. Harry Brignull, who coined "dark patterns," now favours "deceptive patterns", partly because "dark" carries unhelpful connotations and partly because "deceptive" names the actual harm. Regulators use both. Use whichever your team finds clearer; the catalogue of tricks is identical.

My competitors all do this. Won't honest design just lose?

Sometimes, in the short run, on the one metric you're watching. But you're now also competing with regulators and review platforms. The FTC and the EU's DSA have made the hard-to-cancel roach motel specifically actionable, and "I couldn't cancel" is one of the most viral complaints a brand can earn. Honest design trades a small near-term conversion cost for lower churn, fewer refunds and less legal exposure, a trade that compounds in your favour.

How do I spot dark patterns in my own product when I'm too close to it?

Use the five families from the Gray et al. taxonomy as a literal checklist, nagging, obstruction, sneaking, interface interference, forced action, and walk your sign-up, checkout and cancel flows screen by screen. Better still, watch a real person who's never seen the product try to cancel it. The patterns you've normalised will jump out the moment a stranger gets stuck.

Who should own "is this ethical?" on a team?

Make it an explicit role in the design review, not a vibe. Someone, designer, PM, or a rotating "user's advocate", holds the standing brief to ask "whose interest does this serve?" and has the authority to flag a dark pattern before launch. Ethics with no owner is ethics no one defends when the conversion target is missed.

Related in the Toolkit

Where to go next