The Label Paradox: Can AI Transparency Create More Distrust? | AI, Media, and Democracy Lab

Our hopes are pinned on AI labels. AI labels are expected to help users navigate an increasingly synthetic information environment by distinguishing fact from fiction, human from machine. Beginning in August 2025, the EU AI Act requires professional users of GenAI, including journalists, advertisers, political actors, governments, and public institutions, to disclose AI-generated content. The EU AI Office has already proposed concrete approaches for what such labels should look like.

With these labels becoming governance mechanisms of trust online, some crucial questions remain largely unanswered – what if labels do not work as intended? What if they produce unintended consequences, or even undermine the very conditions of trust they are designed to protect?

While public support for AI transparency is high, and mandatory AI labels are an important step towards more transparency on AI for users, we still know surprisingly little about how labels shape user perceptions, platform behavior, and information ecosystems. Does a “Generated with AI” label erode trust in content that is actually accurate and reliable? Does the absence of a label become a proxy for authenticity, regardless of whether content is actually trustworthy? How do users interpret or strategically mobilize labels in politically polarized environments? And could mandatory labeling create a two-tier information ecosystem in which labeled content is dismissed while unlabeled content receives unearned credibility?

These questions become even more pressing in contexts such as journalism, elections, activism, war reporting, and platform moderation, where trust and authenticity are already highly contested. Our online panel brings together experts in AI regulation, media studies, cognitive psychology, computational linguistics, and platform governance to explore the promises, limits, and unintended consequences of AI labeling. Together, we will zoom out and investigate AI labels and their implications as part of the broader information ecosystem. Which problems exactly can AI labels solve, which not, and when do they backfire.

Click here to learn more and register to attend. Note: This webinar is taking place in Central European Summer Time (UTC+2). Event details below are listed in ET (GMT-5).