Poland's UOKiK Tests AI Tool to Hunt Dark Patterns in Online Stores

Poland's Office of Competition and Consumer Protection (UOKiK) has built an artificial intelligence tool designed to independently detect dark patterns, the manipulative mechanisms online stores use to pressure purchases or make it hard to cancel services. The project, funded with EU money, has now entered the testing and results-presentation phase of its first stage.

How the Tool Works

Instead of classic scripts that scan a site's code for predefined patterns, UOKiK opted for a vision-language model that analyzes a website the way a human would. The system evaluates colors, graphic layout, button size and their placement relative to text, letting it catch subtler forms of manipulation than just obvious tricks buried in the code.

In the first stage of the project, the agency's team analyzed more than 300 websites, mostly e-commerce stores, and subjected a hundred of them, selected as using dark patterns, to detailed source code analysis. The material was supplemented with consumer complaints, neuromarketing research using eye tracking, facial expression tracking and EEG, and surveys simulating real shopping scenarios.

What Are Dark Patterns

Dark patterns are interface design techniques that deliberately mislead consumers or make it harder for them to make an informed decision. Common examples include hidden costs revealed only at the final stage of checkout, artificial time pressure created through fake sale countdown timers, unsubscribe buttons tucked away in hard-to-find corners of the interface, and marketing consent boxes checked by default.

UOKiK had previously published a consumer guide titled "Nie klikaj w ciemno!" ("Don't Click in the Dark!"), explaining the most common manipulation mechanisms used by online stores. The AI tool is meant to be the next step, this time aimed not at consumers but at the agency itself as a market-monitoring instrument.

The Official Stays in the Loop

UOKiK President Tomasz Chróstny stresses that automation is meant to support, not replace, the agency's work. The system is intended to speed up detection of unfair practices and catch recurring patterns of manipulation at a scale manual analysis of individual sites could never achieve.

AI will not replace us, or do our job for us, as the guardian of consumer rights in the digital world - Tomasz Chróstny, President of UOKiK

The agency also says it will draw up its own guidelines on the use of artificial intelligence in its operations, meant to prevent a situation where administrative decisions get made automatically, without an official responsible for the final assessment.

Stakes for Polish Online Stores

For the e-commerce industry, this means a real shift in the scale of oversight. Until now, detecting dark patterns relied mainly on consumer complaints and spot checks, which let many stores run manipulative mechanisms for years without consequences. An AI-based tool could let the agency scan the market faster and more systematically, raising the odds that unfair practices get caught even at smaller sellers who have rarely come under the regulator's scrutiny until now.

The project fits into a broader trend of European market oversight bodies turning to artificial intelligence, increasingly relying on automation to monitor large numbers of websites at once. The March ICPEN conference in Bydgoszcz is expected to be an opportunity to trade notes with UOKiK's counterparts from other countries testing similar solutions.

For now, the agency has not disclosed the project's exact budget or a target date for rolling it out fully into inspectors' daily work. The next step is expected to be publication of internal guidelines governing how AI is used in the agency's proceedings.