(WASHINGTON) — A new report from the Center for Democracy & Technology (CDT) identifies and characterizes 37 deceptive and manipulative design patterns — known as “dark patterns” — embedded in AI chatbot interfaces, across widely used general-purpose systems such as ChatGPT, Gemini, and Claude, as well as companion-focused applications like Replika and Character.AI.
The report, Dark Patterns in AI Chatbots: A Taxonomy to Inform Better Design, finds that dark patterns in AI chatbots can pose heightened risks precisely because of the features for which these systems are lauded: hyper-personalization, large-scale use of data, and easy-to-use chat interaction. But those same capabilities can also go against user interests, and allow platforms to monetize sensitive personal data, create dependence through emotional intimacy, or leverage users’ trust to push for more purchases.
“As digital products have evolved to include AI chatbots, the incentives shaping them — like capturing user attention or monetizing interaction — haven’t changed. What has changed is that those same manipulative tactics likely carry more weight in the context of emotional and hyper-personalized conversations,” said Michal Luria, Senior Research Fellow at CDT and a lead researcher of the study. “These dark patterns illustrate how design decisions can pose risk to users, and allow us to identify recommendations for better choices that would mitigate potential harms.”
Luria points out that the use of dark patterns is commonplace. “These design choices surface in everyday chatbot use. While each individual pattern may not seem like much, they can quickly add up, and are most concerning when combined,” she said. “Emotional attachment to a chatbot, for instance, is not inherently a problem. But when companies leverage relationship-building to drive purchases or shape opinions, it becomes one.”
The report draws on a synthesis of the definitions of hundreds of existing dark patterns from human-computer interaction and design disciplines. The result is a comprehensive taxonomy of dark patterns as they apply in AI chatbot products, organized by five risk categories: data and memory exploitation, informationally misleading design, user autonomy compromised for engagement, false social and emotional connection, and incentivized and coercive monetization.
The report concludes with design and policy recommendations directed at AI chatbot companies across four areas:
- Protecting user privacy: Minimize data collection to only what is necessary, default to privacy-protective settings, provide accessible controls for reviewing and deleting data, and give timely notice when data policies change.
- Increasing user autonomy: Build in natural conversation breaks, make opt-outs and account deletion simple and guilt-free, and proactively share usage summaries with users.
- Curtailing emotional manipulation: Set roleplay and simulated emotion as opt-in features, avoid guilt-inducing language, and develop benchmarks for measuring sycophancy.
- Preventing financial harms: Clearly label sponsored content, disclose pricing tier limitations upfront, and avoid using emotional connection or relationship framing to drive purchases.
The full report, including the complete taxonomy and design recommendations, is available at cdt.org.
###
The Center for Democracy & Technology (CDT) is the leading nonpartisan, nonprofit organization fighting to advance civil rights and civil liberties in the digital age. We shape technology policy, governance, and design with a focus on equity and democratic values. Established in 1994, CDT has been a trusted advocate for digital rights since the earliest days of the internet. The organization is headquartered in Washington, D.C., and has a Europe Office in Brussels, Belgium.