The Perils of the “zero click” Internet: How Advertising in AI Models Threatens Free Markets and Free Speech
What happens when advertising no longer directs traffic to a website, but becomes part of the answer itself?
Conversational AI is changing the basic architecture of the internet from a marketplace of clicks into what others have called the “zero-click internet.” The basic idea is that, instead of directing users (via monetized clicks) toward multiple websites, AI models increasingly aim to keep all relevant (monetized) clicks within a single environment, paving the way for AI agents to conduct commerce within a corporate-controlled environment.
There is an important difference between these two models for the Internet:
Search engines provide a ranked set of links, allowing users to compare sources while distinguishing regular results from sponsored placements.
Conversational AI provides a synthesized answer, keeping the user inside an environment controlled by the company operating the AI model.
One big problem: marketers paying for advertising inside AI environments do not pay for a banner or sponsored link; instead, marketers could potentially influence what the AI recommends during the conversation. The commercial intervention moves from buying visibility around the answer to influencing the answer itself.
Our latest commentary in the Journal of Public Policy & Marketing argues that this creates three important risks:
Information exclusion: A single authoritative-looking answer can compress competing perspectives and make alternative viewpoints less visible.
Economic pay-to-play: Large firms may be able to purchase preferential pathways to visibility, making it harder for smaller firms and new entrants to compete.
Social alienation: When AI systems operate as companions as well as commercial products, persuasion can become entangled with trust and human-like relationships.
The problem becomes more significant because conversational AI reduces friction. Users no longer need to open several links, compare sources, or decide which information appears credible. A fluent and confident answer can simply appear to be “the” answer rather than “an” answer. This also creates difficulties for conventional responses to misinformation: fact-checking and media literacy were developed largely for public claims and identifiable sources, not individualized conversations generated privately for each user.
This connects with my broader research on market-oriented disinformation. If the companies controlling AI models can determine which perspectives are served, while commercial or political interests can steer the resulting answers, conversational AI could create a new kind of echo chamber: one generated not primarily through selective exposure between users, but through the architecture of the AI system itself.
In other words, the zero-click internet risks turning AI models into a mass-persuasion system for hire. The policy challenge is therefore not simply whether advertising should exist in AI, but to prevent commercial interests from influencing conversations. We argue for a legally enforceable boundary between sponsored content and the technical output of AI models.
Read the commentary for free:
Diaz Ruiz, C., Arsel, Z., Mas, E. M., & Turner, B. (2026). Advertising in AI Models: How the Zero-Click Internet Threatens Free Markets and Free Speech. Journal of Public Policy & Marketing, 45(3), 307–309. DOI: 10.1177/07439156261436399.
