Remember when the last time was that you have subscribed to an application (for example a calory counting application for myself). All with a great discount and with one simple click. The monthly cost was something like a bottle of water. After some time, you decided that you don’t need that subscription anymore and started to investigate the options to unsubscribe. It took you about 2 days with your busy schedule to figure out the sequence of steps and emails you needed to write and wait till the subscription will be cancelled. Here is the plot – before you managed to cancel yet another subscription fee was charged from your card.
Many of us have been in a situation where we have subscribed to something online with just one click be it music app, cloud expansion, delivery, social media, dating applications or even a gym subscription. In the meantime, when one tries to cancel it, it takes 4 pages, 6 clicks and 8 interactions, 15 options, each offering a discount, a reminder of benefits, or a pause instead of an exit or cancellation. Nothing on those pages is untrue. Nothing is hidden.
The design simply costs more effort to leave than to join — and that asymmetry, deliberately engineered, is what the law now calls a dark pattern.
AI-enabled dark patterns are not just regulatory risk; they are scalable damages exposure. For Armenian e-commerce businesses, this means user-interface choices, pricing logic, and AI personalisation should now be treated as board-level liability issues, not product-optimization details.
Armenia’s sweeping 2025–2026 reforms reshaped its competition and consumer-protection framework, folding both mandates into a single regulator. “Dark patterns” are interface and design choices that push users toward decisions they would not otherwise make. Now in the age of artificial intelligence these patterns have become more powerful, more personalised, and what is crucial - more provable. That last quality is what can convert a design choice into a compensable consumer claim.
Dark Patterns and AI Calculus Change
Dark patterns include familiar tricks: (i) manufactured urgency and false scarcity, (ii) confirm-shaming, (iii) pre-ticked boxes, (iv) hidden or drip pricing, (v) “roach-motel” flows or Hotel California effect that make cancelling far harder than subscribing, and (vi) consent interfaces engineered to nudge a single outcome. Artificial intelligence can supercharge each of those. AI enables hyper-nudging — manipulation personalised to each user, dynamically adapted to their hesitations and weaknesses; it lets firms to test and optimise persuasion at scale, and it powers algorithmic personalised pricing. Such conduct is harder for a regulator or an ordinary consumer to spot — yet the AI systems that generate it log almost everything they do.
Consumer Protection
Armenian law now defines “influence on consumer behaviour” as conduct by which an undertaking directly or indirectly affects (or may affect) the consumer’s informed decision-making, inducing a transactional decision the consumer would not otherwise have made. A parallel concept, “undue influence,” reaches aggressive designs that pressure the consumer even without any physical force. Comparatively the law guarantees free choice without any pressure or restriction, requires personalised pricing to be disclosed, and provides that an order with obligation to pay button which is not clearly labelled leaves the consumer owing nothing. Under the competition legislative framework, all of this qualifies as a practice harming consumers’ interests which the Commission assesses and can prohibit — and, where a rival is targeted, it may equally constitute unfair competition. The revised framework therefore gives the Commission a clearer basis to treat manipulative interface design as consumer harm, while the consumer-law remedies supply the compensation pathway.
.
The Damage-Claims Exposure
Regulatory fines are not the principal risk. Beyond the Commission’s power to order cessation, impose penalties and even apply interim measures during an investigation, the law gives consumers a direct route to full compensation. Where misleading or improper information drives a transaction, the consumer may rescind and recover damages in full, and that right extends to any injured person, regardless of contractual privity. Artificial intelligence multiplies this exposure in three ways.
- The scale: a personalised pattern is deployed to an indefinite circle of consumers, each a potential claimant, and a Commission decision becomes powerful follow-on evidence in their private suits.
- The provability: the logs, experiments and conversion metrics an AI system generates may evidence the design, targeting, and optimisation of the manipulation; causation will still require showing that the consumer’s transactional decision was materially affected.
- The burden-shifting: on several key questions the undertaking — not the consumer — must prove compliance, and deliberately targeting vulnerable users only aggravates liability.
Risk-Control Checklist
Businesses should manage this exposure at three points:
- before deployment, by auditing AI personalisation, consent flows, pricing disclosures, and cancellation journeys;
- during investigation, by preserving evidence and managing Commission engagement; and
- after any finding, by preparing for follow-on compensation claims and, where appropriate, remedies against unjustified measures Each of the above steps is drastically crucial and therefore to harvest the maximum benefit, those shall be performed in cooperation and engagement with experienced team.
.
Conclusion
Dark patterns are no longer a mere design-ethics debate; in Armenia they are an enforcement category with a compensation tail. Artificial intelligence makes manipulation cheaper and more effective, but it also makes it visible and provable — shifting the real danger from regulatory penalties toward private damage claims brought at scale.
The practical takeaway is clear: AI-driven interface design should be governed, tested, and documented as a legal-risk function before it becomes evidence in someone else’s claim.
_____________________________________________________________________________________________________________________________________________________________________________
Author:
Arthur Buduryan
Partner, Legelata Legal and Tax
DISCLAIMER:
This material is produced by or for Legelata LLC. The information contained in this piece is provided for general informational purposes only and does not contain a comprehensive analysis of each item described. Prior to undertaking (or omitting) any action, the reader is advised to seek professional advice tailored to their specific situation. Neither Legelata nor the author accept and hold liability for acts or omissions taken in reliance upon the contents in this material.
LEGELATA LLC 18/09/2026
