A new arXiv preprint reviews more than 15,000 user comments across 59 AI healthcare chatbot apps and finds recurring breakdowns in access and reliability, user experience, billing and support, with privacy and security concerns linked to the worst experiences.

A new preprint on arXiv says AI healthcare chatbots are already functioning like information infrastructure for many users, but that infrastructure is showing familiar failures.

The study, titled AI Healthcare Chatbots as Information Infrastructure: A Large-Scale Study of User-Reported Breakdowns, analyzes more than 15,000 user reviews from 59 AI healthcare chatbot apps. The authors say the feedback points to recurring problems in access and service reliability, user experience and interaction quality, and billing and customer support.

They also say privacy and security concerns are tied to the most negative experiences.

What the study found

According to the preprint, the three main breakdown areas were:

  • access barriers and unreliable service
  • user experience and interaction problems
  • billing and customer support issues

The paper says these failures matter because people are turning to these tools for health information and self-management. The authors frame the apps not just as consumer products, but as part of the information systems people rely on when they want quick guidance or help navigating care.

Why it matters

Healthcare chatbots can be a first stop for users who want answers outside a clinic visit. That makes trust, reliability and privacy especially important.

The authors' findings suggest that problems in basic service quality can quickly become problems of confidence and use. Billing disputes, poor support and privacy concerns can make a tool harder to trust even if the underlying chatbot is technically capable.

Background

The new paper builds on earlier work from the same research group. A separate 2025 preprint focused on privacy in AI healthcare chatbot apps and reported gaps in privacy policies and user control across 12 apps.

That earlier work provides context for the new study's privacy findings, but it is a separate paper.

What happens next

The preprint is very recent, having been posted on arXiv on June 25, 2026. The next developments to watch are whether it is picked up by an institution, appears in a published version, or draws responses from app developers or platform operators.

For now, the paper adds a large-scale view of how users are experiencing healthcare chatbots in practice, and where those systems are failing them.

Revision note

Initial automated publication.