Most people have experienced a moment when something feels a little strange, usually when attempting to decline a newsletter or cancel a subscription. The “Yes, sign me up” button is larger than the “No thanks” button. Three screens down, the opt-out option is hidden. And the cookie rejection link? Accepting everything takes one click, but it takes seven. These are not mishaps. They were intentionally designed this way.
Harry Brignull, a British UX researcher, first used the term “dark patterns” in 2010 after observing that digital interfaces were being purposefully designed to work against users rather than for them. The name stuck after Brignull presented it as a play on “design patterns“—the accepted vocabulary of user interface design. Since then, the field has grown into a wide and unsettling taxonomy of digital manipulation, encompassing everything from outright obstruction to emotionally charged language and visual tricks. Although the term “deceptive patterns” is now preferred by researchers, both terms refer to the same unsettling reality: many decisions that appear to be design choices are actually traps.
It is difficult to ignore the data. Over 10% of a sample of 11,000 well-known e-commerce websites had deceptive design patterns, according to a 2019 study from Princeton University and the University of Chicago. A different University of Zurich study that looked at free trending apps on the Google Play store discovered misleading patterns in 95% of the 240 apps that were sampled, with over half of them having an average of seven such patterns per app. Seven. Sitting quietly in a research paper, that figure reveals a lot about how commonplace this has become.
The modern web relies heavily on A/B testing, which contributes to the rapid spread of these patterns. Businesses test two iterations of a page, retain the version that receives more clicks or conversions, and repeat the process indefinitely. The issue is that a design that deceives someone into unintentionally subscribing also “converts” because the algorithm is unable to distinguish between real interest and artificial confusion. A manipulative interface element can spread from one major platform to every imitator in a matter of months when you factor in the industry’s propensity to copy rivals.
Some of the more pernicious examples are those that initially appear to be almost courteous. Consider confirmshaming, which is when opt-out links are written as “No thanks, I don’t like saving money” or “I prefer to pay full price.” In theory, the design gives you an option. It’s also subtly making fun of you for thinking about it. Another pattern that researchers frequently identify is nagging: applications that repeatedly request permissions for the camera or notifications in the hopes that the user will eventually give up and simply click “Allow” to stop them. Then there’s sneaking, which relies on the fact that most customers move quickly through purchase flows and won’t notice until they’re staring at a receipt. Preselected checkboxes add premium delivery or travel insurance to a cart.
The researcher’s response is truly intriguing because it transcends academic documentation and enters the realm of law. Certain types of consent manipulation are now prohibited by the EU’s General Data Protection Regulation; processing a user’s data without their express consent now has serious repercussions. Amazon’s own subscription cancellation process, which required navigating several screens to get out of Prime, was one of the reasons the U.S. Federal Trade Commission filed a lawsuit against the company. Regulators in a number of nations appear to be taking note of that case, which served as a warning.
Whether legislation will advance quickly enough to keep up with these patterns’ rapid evolution is still up in the air. Product teams chasing conversion metrics and executives who see the research but also the quarterly numbers put pressure on designers who wish to resist. It’s also not always simple to tell the difference between a manipulative design and one that is truly persuasive. A countdown timer and a crossed-out price could be real or fake. Social proof may be staged or it may represent actual user behavior. The distinction between deception and persuasion is real, but it’s not very clear.
Researchers like Brignull and the groups at Princeton, Chicago, and Zurich are working to draw attention to that boundary and make it more difficult to cross it carelessly. The majority of users still don’t have a term for the feeling that an app is subtly working against them, the work is sluggish, and regulations are implemented unevenly. However, that is beginning to change. It becomes more difficult to overlook the issue when the language is present.
