Why “97% of popular sites” does not mean 97% were unlawful

The number is memorable. The qualification matters more.

“Ninety-seven per cent” is the sort of statistic that escapes from a report and begins a life of its own. In this case, it came from a large Commission-commissioned project about dark patterns and manipulative personalisation. Researchers examined popular websites and apps used by European consumers and found at least one dark pattern, under their analytical taxonomy, on almost every service in the mystery-shopping sample.

That tells us something important: these techniques were not confined to obscure corners of the web. They appeared in mainstream digital journeys and across business sizes. It also gives product teams a practical list of places to look first. Hidden information, unequal visual hierarchy, preselected options, repeated prompts, cancellation friction and forced registration are journey features that can be recorded and compared.

But the statistic answers a prevalence question, not a legal one. Researchers were classifying interfaces for a study. They were not issuing 97 infringement decisions. A legal conclusion still depends on the instrument, its scope, the complete interface and operating facts, the audience and the likely effect.

What the project actually combined

The report is more substantial than a simple website count. It brought together market screening, consumer-facing experiments, surveys, interviews and legal analysis. That combination helps answer different questions: what researchers saw, how people responded in controlled settings, how stakeholders described the market and where the existing legal framework might face difficulty.

Those evidence streams should not be collapsed. An observation can show that a design mechanism is common. An experiment can estimate an effect under particular conditions. A legal analysis can identify possible provisions. None of those, alone, proves the outcome of a specific enforcement case.

Why product teams should care now

The study pre-dates any future Digital Fairness Act text, but its operational lesson does not depend on predicting that text. The mechanisms it describes are visible in journeys that already need review under current consumer, platform, privacy and sector rules.

A sensible review therefore begins with the product state. Record the choice as presented, the information available at that moment, the default, the route to the alternative and what happens after the click. Then connect the evidence to the relevant current-law question. The current-law map keeps that legal layer separate, while the visual pattern library makes the mechanisms easier to recognise.

The strongest use of the study

Use the report as a prioritisation tool. It helps identify recurring mechanisms, high-risk journey stages and questions worth testing with users. Do not use it as a shortcut that turns every taxonomy match into an offence. That distinction is not caution for its own sake; it is what makes the evidence useful to both designers and lawyers.

Source check on 14 September 2026

The Publications Office record continues to attribute the 97% result to the mystery-shopping exercise in its 2022 final report. This update does not refresh the underlying observations. The displayed 3% is an arithmetic complement, not a separately verified compliance category or a count of services cleared by an authority.

Questions for a journey review

1. Which of the study’s frequently observed mechanisms appears across our current signup, pricing, subscription and cancellation journeys?
2. Can we preserve the complete sequence and operating state rather than relying on a single screenshot or taxonomy label?
3. Which current-law provision, provider role and likely effect would need separate legal analysis for the observed journey?