Reframing health information quality on social media: An AI governance perspective

Dear Editor,

The study by Chen et al. ¹ offers a timely and necessary look at how osteosarcoma is represented on short-form video platforms such as TikTok and Bilibili. What stands out most is not simply that the overall quality of content is low, but that quality appears to operate independently of visibility, high engagement does not translate to reliable information. This disconnect is not surprising, but it is deeply consequential.

From an information systems standpoint, this is less a content problem and more a design problem. Platforms are functioning exactly as they were built to: optimizing for attention. The issue is that attention is being treated as a proxy for value, even in contexts, such as health, where this assumption does not hold. Similar patterns have been observed in prior analyses of TikTok health content, where engagement does not reliably reflect informational quality.^2,3

What Chen et al. surface empirically is something many of us working at the intersection of AI, education, and digital systems have been observing in practice: systems that are technically efficient but epistemically weak. When recommender systems privilege engagement signals like watch time and interaction, they inadvertently elevate content that is emotionally resonant, simplified, or even misleading. This dynamic aligns with broader concerns around algorithmic filtering and selective exposure, where users are repeatedly presented with content that reinforces engagement rather than accuracy. ⁴

This creates a structural imbalance. Even when healthcare professionals produce higher-quality content, as the study clearly demonstrates, they are competing within an ecosystem that does not reward accuracy in any meaningful way. In effect, credibility becomes secondary to performativity.

The implications go beyond misinformation. In many contexts, particularly in the Global South where access to formal health information systems may be uneven, platforms like TikTok are not just supplementary sources, they are often the first point of contact. In the Philippines and similar settings, this reality means that algorithmic visibility can quietly shape public understanding of disease, risk, and care pathways. When these systems privilege engagement over accuracy, the consequences extend from information gaps to real-world health decisions. This reinforces ongoing concerns in digital health about ensuring that AI systems are aligned with meaningful outcomes rather than purely technical performance.^5,6

Addressing this cannot rely solely on asking healthcare professionals to “create better content.” The burden should not be on individuals to outcompete algorithms. Instead, the conversation needs to shift toward how these systems are designed and governed.

At minimum, this means rethinking ranking mechanisms to account for credibility, completeness, and source authority, not just engagement. It also calls for more deliberate integration of expert knowledge into platform ecosystems, whether through verification layers, hybrid moderation models, or AI-assisted quality assessment. These directions are consistent with broader calls for ethical and accountable AI governance. ⁷

There is also a role for education. As professionals working within higher education, we see a growing need to move beyond digital literacy toward algorithmic literacy, helping learners understand not just how to access information, but how that information is curated, filtered, and prioritized.

Ultimately, the contribution of Chen et al. lies in making visible a problem that is often discussed abstractly. Their findings point to a deeper tension within digital health: we are using systems designed for engagement to deliver information that requires trust.

If digital health is to be taken seriously as a public good, then the systems that distribute health information must be designed to value truth, not just attention.