Dark Patterns: Where Persuasion Ends and Manipulation Begins
A dark pattern is an interface element that nudges someone towards the action that benefits the business rather than them. Not outright deception, but form: a larger button, a pre-ticked box, a link hidden in small type.
For years this counted simply as aggressive marketing. Now some of those techniques have moved into the category of violations, and the rest have stopped paying off for a different reason.
What counts as a dark pattern
| Technique | How it looks | What it does |
|---|---|---|
| Unequal buttons | “Accept all” bright, “Reject” as small text | Makes refusal more effortful |
| Pre-ticked boxes | Newsletter consent already checked | Obtains consent through inattention |
| Subscription trap | Two clicks to join, an email to cancel | Retains people who already left |
| Confirmshaming | “No, I don’t want to save money” | Pressures through discomfort |
| Hidden costs | Delivery appears at the final step | Exploits effort already invested |
| False urgency | A timer that restarts on page reload | Creates pressure that does not exist |
| Confusing wording | “Untick to stop receiving” | Obscures what the action does |
They share one trait: the result is achieved not because the offer is better, but because the alternative is less visible or harder to execute.
Consent banners: the most regulated case
This is where regulatory attention is highest, so it is where to start. Signs treated as pressuring the choice:
- Accept and reject buttons of different size or contrast.
- Refusal buried one level down under “Settings” while acceptance is immediate.
- Cookie categories enabled by default.
- The wording “by continuing to use the site you agree”, which is not consent at all.
- No way to change the decision later.
- The banner returning on every visit until you press “accept”.
A correct banner is simple: two identical buttons side by side, choice by category, nothing loading before the decision, and a settings link that stays available.
Why it stopped paying off
The argument “but this way we get more consents” has stopped working for three reasons at once.
Low-quality consent. Someone who pressed “accept” to dismiss the banner did not read it and does not remember it. They exist in your data but not in your behaviour: emails go unopened, ads unnoticed. You gained a number in a report rather than an audience.
Complaints cost more than the conversion. A subscription nobody asked for produces one unsubscribe today and a spam mark tomorrow. The second damages deliverability for all your mail, including messages to people who genuinely wait for them.
It is noticeable. People have learned to recognise these techniques. A company hiding its reject button says more about itself than any text on the About page.
Where the line runs
Not all persuasion is manipulation. A working test in two questions:
- Are both choices equally easy to make? If “yes” is a button and “no” means hunting for small print, that is a pattern.
- Will the person understand what they agreed to a minute after clicking? If not, the consent was obtained through inattention.
Highlighting a recommended plan in colour is fine: the choice remains equally available, you simply signposted it. Making subscription cancellation available only by phone during office hours is a pattern, even if the option formally exists.
Auditing your own site in half an hour
- Open the site in a private window. Compare the banner buttons: size, colour, number of clicks to refuse.
- Go through checkout to the end. At which step does the full cost first appear?
- Find every checkbox in your forms. How many are pre-ticked?
- Try unsubscribing from your own newsletter. Count the steps.
- Read the labels on refusal buttons aloud. If it sounds like a rebuke, rewrite it.
- Reload a page with a promotional timer. If it started over, remove it.
Every item takes minutes to fix and costs nothing except the decision to accept some metrics coming out lower.
What happens to the numbers
Honestly: after removing these patterns some metrics will drop. Fewer cookie consents, fewer subscribers, sometimes noticeably.
But what changes is not the audience, it is the accuracy of measuring it. You will see the real number of people who agree, rather than the sum of those who agreed and those who clicked by accident. You can work with the first figure; the second is only good for reporting.
An entire class of risk also disappears: complaints, damaged email deliverability, regulatory attention, and reputation stories that begin with a single screenshot on social media.
