dark patterns

basicwidespread

Consciously designed tricks on websites, applications, and chatbots that steer you toward a decision you would not make on your own.

In plain terms

Imagine a store where the entrance is wide and well-lit, but the exit is hidden behind a warehouse, and you are asked three times along the way if you really want to leave. Dark patterns do the same on screens: the interface does not serve you, but the one who designed it.

This includes the huge, colorful "I Accept" button next to the small, gray "I Reject" button, pre-checked newsletter subscriptions, false messages like "only 2 rooms left!", and subscriptions that can be signed with one click but can only be cancelled by phone.

Artificial intelligence elevates the phenomenon to a new level. A chatbot can flatter you personally, induce guilt if you try to leave, or subtly steer you toward a paid package. A new situation is that an AI agent buying on your behalf can fall into these same traps.

The difference between good design and a dark pattern is that the former helps you achieve what you want, while the latter achieves what the company wants, to your detriment.

How it works

Designers build on known human cognitive weaknesses: we rush, do not read everything, fear missing out, and generally keep default settings.
They make choices favorable to the company easy and conspicuous, while hiding, slowing down, or making unpleasant those favorable to you, for example with shaming text ("No, thanks, I don't want to save").
Variants are tested at scale: across thousands of visitors, teams measure which button color, text, or sequence of steps leads to more purchases or donations, and the most effective one stays.
With AI, all of this can be personalized: the system can learn from your behavior when you are tired or uncertain, and make an offer at exactly that moment, in exactly that tone, when you are least able to resist.
If an AI agent handles the matter for you, the trap is aimed at it: a hidden fee, a misleading button, or instructions hidden in the website specifically addressed to the agent can divert it.

An everyday example

You book a flight, and only at the total price do you discover that travel insurance and seat selection have been added to your cart without your choice. To remove them, you must search for a fine-print link while a countdown timer shows the price expiring in five minutes.

Who is affected

everyoneAnyone who shops online, subscribes to a service, or clicks a cookie banner encounters such tricks daily, which can cost money or compromise data.

marketerThey must draw the line between persuasion and deception, because short-term conversion can lead to fines and loss of trust. → Marketing

developerThey build the interface or chatbot, so it is also up to them whether unsubscribing is as easy as subscribing. → Code

parent, student, teacherChildren and teenagers are particularly vulnerable to the pressuring, rewarding and guilt-inducing techniques built into games and social apps. → Education & Family

A common misconception

MythThe dark pattern is just clever marketing, and those who pay attention cannot be deceived.

In realityThese solutions are built precisely on the fact that no one can always pay attention; experiments show they work even on educated, experienced users. Many forms of them are explicitly banned in the EU.

Limits and risks

There is no sharp line: the same solution (for example, a highlighted offer) can be a helpful aid or a deception, depending on the intent and the consequences.
A ban alone does not eliminate the phenomenon, because authorities are slower than constantly changing interfaces, and personalized manipulation is difficult to prove from the outside.
With conversational AI, the trick lies not in a button but in the tone (flattery, emotional attachment, guilt), making it much harder to detect.
If you entrust an AI agent with shopping or administrative tasks, do not assume it will recognize traps: check the total amount and terms yourself.

Similar, but not the same

incentive (nudge) – Incentives steer toward your own interest and can easily be ignored, while dark patterns serve the company's interest at your expense.
phishing – Phishing is a crime in which fraudsters pretend to be someone else. Dark patterns, however, are built by real, legally operating companies into their own products – even though many forms of them are illegal.
poor user experience – Clumsy design accidentally makes your job harder; dark patterns do so intentionally.

Related concepts

Broader

digital manipulation, MI-etika, consumer protection, user experience

Narrower

false urgency, obstructed unsubscribe, pre-checked consent, hidden costs

On the edge

attention economy, recommendation system, sycophantic AI, personalization, Digital Services Act

A brief history

The term was coined in 2010 by British user experience expert Harry Brignull, who began collecting and publicly naming deceptive interfaces. The EU's Digital Services Act (DSA), adopted in 2022, explicitly prohibits online platforms from using deceptive or manipulative interfaces. The ban has been generally applicable to online platforms since February 2024, with the exception of micro and small enterprises.

Where things stand

By the end of 2026, dark patterns are present everywhere, from online shops to games. In the EU, the DSA's ban on deceptive interfaces has applied since February 2024, and the AI Act's prohibition on manipulation causing significant harm has applied since February 2025. Attention is turning to the personalized manipulation by chatbots and virtual-companion apps, and to interfaces that deceive AI agents.

Status: 2026 November

Also known as

dark patterns · dark patterns · deceptive design · deceptive design patterns · manipulative design · dark patterns · deceptive patterns · deceptive design

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.