New research presented at the American Academy of Pediatrics 2026 National Conference & Exhibition in San Diego offers one of the clearest looks yet at how TikTok is shaping a fragile and anxious public conversation about medication use during pregnancy. The study analyzed 53 short-form videos discussing the alleged relationship between prenatal acetaminophen exposure and autism, a topic that has generated intense online debate after claims circulated that the common pain reliever, sold widely under the brand name Tylenol, causes neurodevelopmental harm. The results were striking: 52.8% of videos contained information consistent with available evidence, 28.3% contained misinformation, and 18.9% were classified as mixed or uncertain. Inter-rater reliability for the misinformation classification was substantial, with a Cohen κ of 0.86, indicating that independent reviewers strongly agreed about which videos were misleading. The research team was motivated by a real clinical problem: pediatricians and other clinicians reported seeing pregnant patients in their clinics asking whether it was safe to take Tylenol during pregnancy, and prior reports suggested that misinformation about acetaminophen and autism had surged on social media. Acetaminophen is one of the most widely used medications for pain and fever during pregnancy, and the scientific consensus is that no causal link has been demonstrated between its use and autism. Yet the findings suggest that TikTok users are encountering a very different message, often delivered by personalities with little medical expertise and packaged in ways designed to generate views rather than accurate health communication. TikTok now functions as a search engine for many adolescents and young adults, who increasingly rely on short-form video rather than traditional medical sources for answers to pressing health questions. For patients of reproductive age, that means the first answer they receive about a medication during pregnancy may come not from a physician but from an influencer, a content creator, or an anonymous account. The new study suggests that the quality of that advice varies dramatically by who is providing it, and that the videos most likely to be seen are not necessarily the ones that are true.
One of the most compelling findings was the sharp divide according to creator type. In the sample, health professionals produced exclusively evidence-consistent content. Not one video from a health professional was classified as misinformation. By contrast, influencers—online personalities who gain large audiences through engagement and personal branding rather than clinical credentials—accounted for 73.3% of the videos classified as misinformation, a difference that was statistically significant at P < .001. This pattern demonstrates that the sources of false claims about acetaminophen and autism are not doctors, nurses, academic medical centers, or public health agencies; rather, the dominant producers of misinformation are content creators whose financial and reputational incentives are tied to clicks, shares, and controversy. Influencers often speak in relatable, direct, and emotional tones, which can make their messages feel more trustworthy than the cautious, qualified statements of medical professionals. The study suggests that this trust is misplaced when it comes to prenatal acetaminophen information. The finding does not mean that all influencers are irresponsible, but it does mean that, in this highly sensitive corner of maternal health, they are the primary source of false and misleading claims. The researchers also noted that the distinction between “evidence-consistent” and “misinformation” mattered for more than just accuracy. Videos containing misinformation also tended to be more extreme in style and more likely to present medical information as a battle between ordinary people and powerful institutions. That framing appears to be especially effective at capturing attention, which raises concerning questions about the incentives of algorithmic platforms that reward engagement. Misinformation videos also scored far lower on educational quality. In the study, misinformation videos had a mean Global Quality Scale score of 1.73, compared with 4.18 for evidence-consistent videos, a difference that was strongly significant at P < .001. The Global Quality Scale is a widely used tool that rates the educational value of online health content on a five-point scale, with higher scores indicating content that is clear, complete, and useful for viewers as a source of health information. A score of 1.73 suggests that misinformation videos were not merely somewhat less educational; they were, on average, poor-to-inadequate in providing meaningful guidance. The evidence-consistent videos, by contrast, approached the higher end of the scale, meaning they were judged to be clear and helpful. Among videos classified as misinformation, 46.7% included conspiracy framing and 53.3% included pharmaceutical distrust. These themes often appeared together, with videos suggesting that drug companies, doctors, or government institutions were hiding the “truth” about Tylenol, or that the fear of autism was being covered up in order to protect pharmaceutical profits. That kind of messaging is deeply powerful because it provides a simple villain and a simple solution: avoid a common medication, trust only the messenger. It also makes the advice almost impossible to rebut with scientific evidence alone, because every rebuttal can be framed as part of the alleged cover-up. The combination of low educational quality, conspiracy themes, and pharmaceutical distrust was a hallmark of the videos that were most likely to be flagged as misinformation by the study’s reviewers. To determine what really predicted misinformation, the researchers performed an adjusted analysis that controlled for multiple factors simultaneously. The results were important for understanding why some videos misled viewers. Lower educational quality was independently associated with misinformation, with a P value of less than .001, meaning that even after accounting for other characteristics, poorly made videos were significantly more likely to contain false claims. Conspiracy framing was also independently associated with misinformation, at P = .012, meaning that videos containing a conspiratorial narrative were significantly more likely to be false, even when educational quality was taken into account. Perhaps most surprisingly, creator type was no longer independently associated with misinformation after adjustment. In other words, the reason influencers were more likely to spread misinformation was not simply that they were influencers; rather, their videos were more likely to be low quality and to use conspiracy framing. Once those characteristics were accounted for, the statistical link between creator type and misinformation disappeared. This is an important insight because it suggests that the problem is not simply a matter of “trained professionals versus online charlatans.” Instead, it is about the characteristics of the message itself. A health professional could, in principle, create a misleading video if it were poorly constructed and relied on conspiratorial themes. Conversely, an influencer could create an accurate video if it were well researched and carefully presented. The finding points to the need to improve educational quality and reduce the appeal of conspiratorial explanations, rather than simply attacking the people who create content. It also highlights a role for media literacy education, so that viewers can recognize low-quality, conspiracy-driven health content, no matter who publishes it. The most concerning finding, perhaps, is the relationship between misinformation and reach. Although misinformation represented slightly more than one-quarter of the sample, it generated more than half of all total views. The difference between the proportion of misinformation videos and their share of viewership was pronounced. A single misinformation video received 10.3 million views, accounting for 36.6% of all exposure across the videos analyzed. That means one false video was responsible for more than a third of the total viewership in the entire sample, a level of exposure that dwarfs many evidence-based videos that attempted to explain why there is no causal evidence linking prenatal acetaminophen exposure to autism. The researchers interpreted this as evidence that TikTok’s algorithm amplified low-quality, conspiracy-framed videos containing misinformation on a very sensitive health topic, allowing them to spread far wider than accurate, balanced content. The research team’s statement was direct: “Our results indicate that TikTok’s algorithm amplified low-quality, conspiracy-framed videos containing misinformation on a very sensitive health topic and allowed for misinformation videos to spread far wider than evidence-based videos that explained the lack of causal evidence between prenatal acetaminophen exposure and autism.” However, the study’s authors were careful to note an important limitation. Because this was a cross-sectional analysis of selected videos and their engagement metrics, the findings demonstrate disproportionate exposure within the sample rather than directly establishing how TikTok’s recommendation algorithm caused the differences in reach. In other words, the study cannot prove that the algorithm actively promoted the misinformation videos; it can only show that, at a single point in time, these videos were receiving far more views than the accurate content. Even with that caveat, the scale of the imbalance is difficult to ignore, and the study raises serious questions for social media platforms, public health agencies, and medical organizations about how to ensure that reliable information is not drowned out by sensational falsehoods. For pediatricians and other clinicians, the study has immediate practical implications. The patients most likely to see misinformation about prenatal acetaminophen are adolescents and young adults of reproductive age, the very population that pediatricians are often responsible for counseling. Social media is now a major source of health information for this group, and the study suggests that the information they encounter may be actively harmful. The lead researcher, Srivastava, described the clinical origin of the project: “After seeing pregnant patients in our own clinics question whether it was safe to take Tylenol during pregnancy, we wanted to understand what information they were being exposed to online. Our findings reveal how important it is for medical institutions to be part of the conversations on social media by sharing accurate, engaging content that reaches patients where they are.” That statement points toward a solution that is not just about debunking myths, but about building a stronger, more credible presence on the platforms where patients actually get information. Medical institutions, pediatric practices, and public health organizations may need to collaborate with social media creators to produce accurate content that is also emotionally compelling and algorithmically visible. They may also need to respond quickly when misinformation emerges, as it did with the claims about acetaminophen and autism. For individual clinicians, the study suggests that asking patients, especially young patients, whether they have seen TikTok videos about their medications may be an important part of routine care. Helping patients understand the difference between a scientifically supported source and a conspiratorial influencer may be as important as the clinical advice itself. Ultimately, the study is a reminder that the fight for accurate health information is no longer limited to clinics, schools, or even websites. It is being waged on short video platforms, where a single misleading clip can reach millions of people, and where the algorithm may reward the loudest voice rather than the most accurate one.
