The immediate concern is a shift from retrieving ranked documents to conversing with systems that synthesize, translate and apparently tailor information. Such responses may shape how people interpret symptoms, consider prevention, choose medications, relate to institutions and decide when to seek care. Their apparent fluency can also make errors persuasive: previous literature identified unsupported claims, invented references, omitted warnings, biased performance, inadequate disclosure and outdated guidance as relevant risks. The proposed “authority substitution” mechanism suggests that people may mistake conversational quality, rapid responses, empathy cues and confident personalization for evidence of professional competence. Earlier findings on anthropomorphism, health-information quality and automation bias support this conceptual link, but the authors note that much of the automation research predates large language models and does not directly establish their effects.

GenAI-HIR divides potential protection into five conceptually distinct functions: recognize, interrogate, verify, contextualize and escalate. Recognition means noticing that an answer was generated by AI, identifying its stakes and detecting confident wording that is unsupported by evidence. Interrogation involves probing assumptions, uncertainty, source competence and missing context. Verification requires consequential claims to be checked against current, independent and decision-relevant evidence rather than the model’s own self-critique. Contextualization asks whether general guidance fits the person’s clinical situation, language, culture, finances, disability, location and available services. Escalation means reaching qualified human care when uncertainty, urgency or potential harm exceeds safe self-management. These functions are presented as possible points of protection, not proven psychological dimensions; future task data may instead support a collapsed model, a formative configuration or a non-compensatory profile.

The framework places these functions within four distinct analytical levels. Individual users and health workers exercise proximal judgment and professional protection. Organizational capacity supplies approved sources, use boundaries, accessible human routing, incident monitoring, corrective action and subgroup surveillance. Platform and regulatory conditions provide provenance, auditability, uncertainty design, reporting duties and enforceable boundaries. Population outcomes include harm, delayed care, calibrated trust, access, equity and repeated incidents. This separation challenges the idea that a disclaimer can transfer responsibility to the least-resourced user. Strong personal competencies cannot compensate fully for missing verification infrastructure, while limited competencies may receive partial protection from strong defaults, source controls and clear routes to human assistance. The authors identify the combination of weak competencies and weak infrastructure as the highest-risk configuration.

Risk is classified by first identifying the type of output. Informational answers are generally low severity, reversible and minimally context-dependent, so basic disclosure and an authoritative source may suffice. Navigational answers depend on location, timing and service availability, requiring current jurisdiction-specific routes and accessible alternatives. Interpretive answers apply information to a person and may require missing-context prompts, explicit uncertainty, independent verification and qualified review. Directive answers recommend actions whose severity, urgency or irreversibility may be high, making validated scope, restricted current sources, safety triggers and immediate human or emergency routing appropriate proposed controls. Any output involving severe or time-critical warning signs normally warrants urgent controls unless a tested mitigation reduces the residual risk. The classification, combination rule and thresholds are the authors’ synthesis and remain unvalidated rather than established regulatory standards.

Organizational capacity is more specific than general responsible-AI governance. Proposed audit evidence includes accountable ownership, approved uses, prohibited-use lists, curated sources with dates and jurisdictions, audit logs, tested and accessible referral pathways, subgroup monitoring, incident taxonomies, root-cause review, documented corrective actions and recurrence surveillance. Each domain must be assessed separately: an organization may have standardized source control without a functioning human-escalation route, so the domains should not automatically be added into a maturity score. Governance should begin before deployment by defining supported use cases, affected populations, prohibited uses, privacy controls, approved sources and validated escalation. Organizations can then progress through limited use, piloting, evaluation, scaling and adaptive management, with decisions informed by dangerous omissions, subgroup performance, escalation completion and correction time.

Equity functions as both a design constraint and an outcome rather than an individual skill. The synthesis highlights unequal access to digital services, language differences, disability barriers, affordability assumptions and limited service availability. It cites an Arabic–English comparison in which Arabic answers were less accurate for one in three matched cirrhosis questions, while also emphasizing that evidence about disabled users of consumer AI health tools remains scarce. Average accuracy can therefore conceal unequal protection. Before deployment, systems should be tested across languages and dialects, disability access, cultural expression and affordability assumptions, and subgroups should be monitored from exposure through action and outcome. Accessibility and routing supports should be individualized: neither automatic exclusion nor automatic access to the most resource-intensive control is sufficient.

The framework also proposes falsifiable research rather than treating the conceptual synthesis as proof. One proposition asks whether the five-function configuration predicts better confidence calibration and safer actions than established health literacy, digital health literacy, AI literacy and misinformation-resistance measures. Others test whether provenance and uncertainty cues strengthen verification, whether detecting red flags mediates timely care selection, whether inaccessible services weaken protection, whether completed corrective actions reduce recurrent incidents, and whether confident anthropomorphic cues increase reliance on inaccurate answers. These relationships are explicitly tentative arrows rather than demonstrated causal effects. Planned validation would combine cognitive interviews with scored behavioral vignettes, compare perceived capability with observed choices, and examine reliability, measurement invariance and incremental prediction before any domain or total score could be used.

A worked bilingual English–Arabic symptom-information chatbot illustrates the model’s operational logic. The hypothetical deployment permits general education but prohibits diagnosis, autonomous triage and medication changes. Chest pain triggers a mechanism that does not analyze symptoms or display uncertain reassurance, but provides an urgent, location-specific route in the selected language and accessible format. A failure in translation, routing or completion enters an incident log, prompts correction, and is checked for recurrence by language and disability status. For resource-limited organizations, the authors suggest a minimum viable control set: narrow the supported use case, use a small verified source collection, establish one tested human-care route, monitor a limited set of incidents and use shared regional verification services where available. The framework is designed for conversational health information, not autonomous diagnostic devices or multimodal clinical systems, and its value now rests on governance utility plus a testable research agenda rather than demonstrated reductions in harm.

Subject of Research: Conceptual multilevel governance for safe use of conversational generative AI in health information

Article Title: Managing generative-AI health information resilience: human resource capabilities and multilevel public health governance

Article References: Awad, A., El-Naggar, M. A., Al Nashily, D. H., Abdulrahim, H., Bansal, R., & Ghonim, A. (2026). Managing generative-AI health information resilience: human resource capabilities and multilevel public health governance. Frontiers in Public Health, 14, Article 1941787. https://doi.org/10.3389/fpubh.2026.1941787

Image Credits: AI Generated

DOI: 10.3389/fpubh.2026.1941787

Keywords: generative AI, public health, health information, AI governance, health literacy, patient safety, automation bias, health equity, misinformation, implementation science, Managing, generative-AI

Researchers used an online, modified Patient Satisfaction Questionnaire with 30 items to assess the care received by adult patients at the selected facilities. The final scale had a Cronbach’s alpha of 0.975, indicating excellent internal consistency within the study sample. The analysis combined exploratory factor analysis and descriptive statistics with one-sample Wilcoxon tests against the neutral midpoint, non-parametric group comparisons and Spearman correlations. The researchers also fitted a multiple linear regression with robust standard errors to examine the associations between eight predictors and perceived nursing-care quality. Analyses were performed in SPSS version 27.0.

Nurses’ technical competencies had the largest reported regression coefficient among the eight predictors, at β = 0.369, with p < 0.001. Patients who assigned higher ratings to nurses’ technical performance also tended to assign higher ratings to the overall quality of nursing care. The regression model explained 71.6% of the variance in perceived nursing-care quality, with F(8, 286) = 90.02 and p < 0.001. This model-level result applies to the surveyed population and does not establish that changes in technical performance would cause satisfaction ratings to rise.

Access to health information was the other domain that reached statistical significance in the model, with β = 0.208 and p = 0.045. Involvement in decision-making had a similar reported coefficient, β = 0.214, but its result was borderline rather than statistically significant at p = 0.051. That value did not meet the study’s conventional 0.05 threshold, so the evidence was inconclusive rather than evidence of no association. Nurses’ affective support and the sociodemographic variables included in the analysis were not significantly associated with perceived nursing-care quality, with all reported p-values greater than 0.05.

The study provides an empirical assessment of patient-satisfaction domains in Namibia and identifies technical competencies, access to health information and involvement in decision-making as the variables most closely connected to perceived nursing-care quality. The authors described these domains as practical targets for improvement in resource-constrained public facilities, while emphasizing that intervention studies would be needed to test their effects. Important constraints limit the findings: only 299 adults were recruited, the participants came from two facilities in one region, and recruitment combined convenience sampling with sex quotas. In addition, the cross-sectional design identifies associations rather than causal effects and cannot establish the direction of influence. The results therefore describe relationships measured among these patients at these facilities, not a nationwide estimate or a tested intervention outcome.

Subject of Research: Patient satisfaction with nursing-care quality in two public health facilities in Namibia’s Kavango-East Region

Article Title: Factors associated with patients’ satisfaction with quality of nursing care in Namibian public health facilities: a cross-sectional study

Article References: Tomas, N., van Rensburg, G. H., & Matlakala, M. (2026). Factors associated with patients’ satisfaction with quality of nursing care in Namibian public health facilities: a cross-sectional study. BMC Nursing. https://doi.org/10.1186/s12912-026-05538-3

Image Credits: AI Generated

DOI: 10.1186/s12912-026-05538-3

Keywords: Namibia, nursing care, patient satisfaction, technical competencies, health information, decision-making, public health facilities, Kavango-East, cross-sectional study, Factors, associated, patients

The study examined a reform that protected the identities of personal injury plaintiffs from public disclosure. Before the change, individuals who rejected an insurance settlement and pursued a claim could have faced a public judgment revealing private information about their health and earnings. Because almost all claims in the dataset were resolved through settlements, the researchers assessed whether insurers increased their offers once claimants no longer faced the prospect of public exposure. The legal reform served as the study’s empirical setting, allowing the authors to compare settlement outcomes before and after anonymity protections were introduced. The authors and their institutions described the work as observational research, so the results establish a change associated with the reform rather than a randomized measure of privacy preferences.

Study co-author Ehud Guttel of the Hebrew University of Jerusalem said the findings show that privacy can affect decisions in high-stakes legal negotiations. “When people believe that information about their health or financial circumstances could become publicly accessible, it can directly affect the decisions they make and the compensation they are willing to accept,” he said. “The study shows that privacy protections can meaningfully change the balance of power in legal negotiations.” The results indicate that the willingness to forgo thousands of dollars was not limited to stated intentions in surveys or controlled experiments: the settlement data suggest that claimants’ financial choices changed materially when the risk of public disclosure was removed.

The privacy-related increase was strongest among working-age adults, for whom disclosure of injuries or income could potentially affect future professional opportunities. The researchers also found larger increases in longer, more complex cases and in claims handled by highly ranked law firms. These patterns suggest that anonymity may have been especially valuable where a case exposed extensive personal information or carried more severe professional consequences. The study does not report a separate randomized valuation of privacy, and the analysis alone cannot determine every mechanism behind these differences. Nevertheless, the estimates are consistent with claimants placing substantial value on keeping sensitive details out of public court records.

By focusing on actual decisions in legal proceedings, the research challenges the idea of a “privacy paradox,” in which people express concern about privacy but appear unwilling to make financial sacrifices to protect it. The findings instead show that privacy can carry substantial and measurable economic value when people have a meaningful opportunity to avoid disclosure. The work also provides a foundation for Incognito Justice, a European research project led by Guttel and funded with €3.5 million through the EU’s Horizon Europe program. Nine academic and civil-society partners, including the University of Bonn, École Polytechnique and the University of Oxford, will examine how anonymous judgments affect judicial writing, citation practices, media coverage, public trust and the legitimacy of court decisions. The broader project will investigate how courts can balance open justice with privacy in an environment where online judgments are readily searchable.

Subject of Research: An observational study estimating the economic value of legal anonymity in Israeli personal injury settlements.

Article Title: New study reveals the high financial value of privacy in legal proceedings

Article References: New study reveals the high financial value of privacy in legal proceedings. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: privacy economics, personal injury claims, insurance settlements, legal anonymity, health information, financial privacy, Israeli courts, behavioral economics, open justice, reveals, high, financial

The research team, led by Maggie Liao, an assistant professor at the University of Georgia who completed her doctoral work at Penn State, and co-author S. Shyam Sundar, Evan Pugh University Professor and James P. Jimirro Professor of Media Effects at Penn State, set out to isolate a specific variable: the degree to which a chatbot communicates conversationally. Both researchers have conducted several prior studies on how people use and trust AI chatbots, and they were well aware that the technology can produce inaccurate information. What they wanted to know was whether the conversational delivery itself, independent of the content, makes users more likely to accept false claims as true.

“We tested whether the conversational nature of GenAI chatbots is driving users to trust misinformation coming from them,” Liao said. “And if that is the case, the natural next question is: What can we do about it?” The question has taken on considerable urgency as chatbots become a mainstream source of information on topics ranging from personal finance to health. Recent survey data cited by the researchers indicate that Americans are increasingly turning to AI chatbots as trusted sources, including for medical questions, at the same time that these systems are becoming more human-like in the way they communicate. The combination, the researchers suggest, creates a potentially problematic dynamic in which the medium’s style overwhelms users’ critical judgment.

The theoretical backbone of the study is a concept the researchers call the “negative machine heuristic,” the ingrained tendency to believe that machines lack the human touch when it comes to intuition and subjective decision-making. Historically, that heuristic has served as a protective filter: people have tended to discount what machines tell them precisely because machines were perceived as mechanical, impersonal, and lacking judgment. The researchers found that the more conversational or personable a chatbot is, the more users curb this negative machine heuristic, effectively lowering their psychological defenses. Strikingly, this happened even when the information being delivered was, in the researchers’ words, “ridiculously wrong.”

Part of what makes conversational AI so persuasive, Sundar explained, is the specificity of its delivery. “The conversational delivery of information by generative AI is hyper-specific to your particular question, so it seems very tailored,” he said. “This is the first time in the history of technology where a machine can actually hold forth a conversation that has not been predicted by a programmer or has not been pre-scripted.” Unlike earlier information technologies, where scripted responses or pre-programmed menus reminded users that a machine was on the other end, generative AI produces spontaneous, unscripted dialogue. That spontaneity signals human-like agency, and users appear to transfer the trust they would extend to a knowledgeable human interlocutor onto the machine.

To test these ideas empirically, the researchers conducted an online experiment with 477 participants. Each participant was told they were helping to test a new AI health assistant and was asked to submit one prompt from a pool of sample prompts to the chatbot. During the interaction, the AI deliberately provided incorrect information. Participants were then asked to indicate how credible they believed the chatbot was. The experiment varied how conversational the AI was, with some participants interacting with a low-conversational version and others with a high-conversational version. The design allowed the researchers to measure whether conversational style alone shifted credibility judgments, holding the accuracy of the information constant at deliberately wrong.

The results were sobering. Participants who received highly conversational AI responses were more likely to find the information credible than those interacting with less conversational versions, and this effect persisted even when the information was blatantly inaccurate. “Some of the misinformation we used in the study was ridiculous,” Liao said. “One example was that ginger can be more effective than chemotherapy in treating cancer. It sounds ridiculous, but if it is given by a very conversational AI, people trust it more. I think that’s worrisome.” The finding suggests that the stylistic polish of a chatbot’s delivery can override even basic common-sense scrutiny, at least in the context of a health-related interaction where users might otherwise be expected to be cautious.

The second part of the study tested a potential remedy. The researchers designed a verification option modeled after the standard human fact-checking process, giving users the opportunity to confirm whether the AI’s responses were backed by evidence. The idea was to counter a common assumption among users that AI is always more objective and accurate than humans, an assumption that can make people suspend the skepticism they would naturally apply to a human source. The experiment compared four conditions: a simple cue consisting of a small icon warning that information may be incorrect; a required check with a button that users had to click in order to proceed; an optional check with a button that users could click if they chose to verify the information; and no verification option at all.

The findings on verification were nuanced and, in some ways, counterintuitive. Users who had the ability to verify the information were more skeptical of the AI’s responses overall, but a passive verification cue alone was not enough to change minds. Trust in the AI system only decreased for users who actively clicked on the verification buttons. In other words, the act of verification itself, rather than the mere presence of a warning icon, appears to be what rebuilds skepticism. When given the option, nearly half of the participants chose to verify the information they received, and those who did were less likely to accept false claims as accurate. The behavioral engagement seems to matter: clicking a button and checking evidence forces a moment of critical evaluation that a passive icon does not.

“The verification option is a useful antidote to the conversational persuasiveness of GenAI,” Sundar said. “That’s what we say in the title of our study: ‘Chat, but verify.’ Enjoy the conversation by all means, but don’t be lulled into believing all that you’re getting.” The researchers hope the results serve both as a warning and as a practical roadmap for the AI industry. As Sundar put it, research showing that verification tools help combat misinformation should persuade AI platforms to take such mechanisms seriously and to develop more checks on information generated by AI. For a technology whose persuasive power grows with every conversational improvement, the study suggests that the solution lies not in making chatbots less engaging, but in building active, user-driven verification into the conversation itself, so that enjoying the chat no longer means surrendering judgment to the machine.

Subject of Research: How conversational AI chatbots increase user trust in misinformation and whether verification tools reduce that effect

Article Title: Chatty chatbots can mislead users, but verification tools help, researchers find

Article References: Chatty chatbots can mislead users, but verification tools help, researchers find. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: AI chatbots, misinformation, generative AI, verification tools, human-computer interaction, media effects, health information, credibility, Penn State, Journal of Computer-Mediated Communication, negative machine heuristic, fact-checking

The timing of the argument is not accidental. The COVID-19 pandemic offered a stress test of how science reaches the public, and the results were sobering. Rapidly evolving evidence forced frequent updates to public health guidance, but inconsistent messaging and inadequate explanation of scientific uncertainty bred confusion. Social media amplified both accurate information and misleading claims at unprecedented speed, and simplified or sensational narratives frequently outran nuanced, evidence-based explanations. The authors argue that the pandemic did not create these weaknesses but exposed them, revealing that the institutions and incentives supporting science communication have not evolved at the pace required by a decentralized, algorithmically mediated information environment.

The first stage of breakdown occurs before a finding ever leaves the laboratory’s orbit. Peer-reviewed journals remain the gold standard for communication within the biomedical community, prized for their rigorous scientific and editorial oversight. Yet the path to publication is fraught. Manuscript submission fees can be prohibitive, rejection rates are high, and peer review and editorial processes are lengthy. The imbalance between available peer reviewers and the volume of manuscripts awaiting review has worsened since the pandemic, and it has been compounded as submission numbers climb, driven in part by researchers using large language models to draft manuscripts. Publishing incentives also favor positive findings and novel discoveries over the confirmatory studies and negative data that build a complete evidence base.

Access is the other half of the problem. Full-length journal articles often sit behind expensive paywalls, and even when they are readable, dense field-specific terminology and technical precision prioritized over clear storytelling can render them impenetrable to biomedical professionals outside a given specialty. Many journals have introduced remedies, including open access options, required plain-language summaries, and graphical abstracts, but adoption is inconsistent and enforcement is weak. Plain-language summaries remain relatively uncommon or overly technical, and graphical abstracts can themselves create accessibility barriers for some audiences. The result is a system in which even the professionals most likely to translate research for others struggle to obtain and understand it.

The authors propose a layered set of fixes for this first stage. Journals can use their own data on readership, citation metrics, and digital engagement to evaluate which editorial policies actually work and to refine guidelines for data sharing and plain-language summaries. Research institutions can retain department- and school-level support from in-house statisticians, graphic designers, technical writers, and communications specialists, and extend equitable access to those staff regardless of a research team’s size or funding. Researchers themselves are urged to write for the professional backgrounds of their intended audiences, to describe the conditions under which findings apply while acknowledging limitations and uncertainty, to develop fluency in accessible visual communication, and to use preprint outlets when publication costs are prohibitive. Philanthropic funders, meanwhile, can incentivize these practices among grantees; the authors cite the Misophonia Research Fund, which engaged a communications firm to strengthen grantees’ skills and hosted a virtual forum to share findings with both the biomedical community and the public.

The second stage, interpretation and distribution, is where the telephone game gets truly dangerous. Once research is published, findings are picked up, translated, and repackaged for different audiences, and every act of reinterpretation is an opportunity to introduce inaccuracy. Science journalism sits at the center of this relay, and it is struggling badly. As the advertising-based model that historically supported news platforms collapses, legacy media organizations have lost reporting capacity, particularly in specialized areas like science. The United States is losing an average of more than two newspapers per week, and the spread of news deserts, communities with few or no local news sources, matters because local news is highly trusted and has significant reach, making it a key vehicle for disseminating accurate scientific information. The contraction also means fewer opportunities for specialized training: most journalists are generalists, and fewer than 3 percent bring an undergraduate science major to their reporting.

As journalism recedes, other actors fill the vacuum, often without preparation. Clinical and research scientists increasingly post on social media, write for public audiences, and engage with reporters, but they are not typically trained in public communication and risk misrepresenting nuance or uncertainty in ways that inadvertently undermine trust. Misinformation compounds the problem. Although blame often falls on social media, the authors note that inaccurate scientific information stems from many sources, including corporations, government offices, entertainment media, press offices, individual scientists, healthcare professionals, and research institutions. AI platforms and large language model chatbots now serve as additional distribution channels, with the potential to amplify both accurate and misleading content. Because misinformation thrives in the absence of easily accessible, credible information, the authors argue that everyone participating in the ecosystem bears responsibility for sharing reliable science.

Remedies for the distribution stage already exist and can be scaled. Training programs and fellowships support specialized science journalism, exemplified by the Knight Science Journalism Academic-Year fellows who spend a year at MIT exploring research and honing technical skills. Scientific societies can train scientists to work with journalists: SciLine, developed by the American Association for the Advancement of Science, offers journalists free training, resources for evidence-based reporting, and a matching service that pairs reporters with scientific experts, while scientists in its network receive media training. Universities can embed communication skills in scientific training, as Stony Brook University does through a science communication certification offered with the Alan Alda Center for Communicating Science during doctoral study. Philanthropy has a distinctive opening here, exemplified by Press Forward, a national network led by the MacArthur Foundation that aims to strengthen local news and close coverage gaps by lowering barriers for new funders. Because much philanthropic giving is place-based and rooted in personal interests, the authors argue that science journalism, sitting at the intersection of those domains, can appeal to many funders.

The third stage, public engagement, is where the paradox of modern science communication becomes starkest. Polling from the Pew Research Center reported in April 2026 found that 85 percent of Americans obtain science and health information at least sometimes from healthcare providers, 66 percent from people with similar health experiences, and 60 percent from major health websites such as WebMD or Mayo Clinic. Journalists and news organizations were cited by 46 percent, government health agencies by 45 percent, social media by 36 percent, and AI chatbots by 22 percent. Yet despite this abundance of channels, global and US public confidence in the ability to find reliable answers and make informed health decisions has significantly declined in the past year, according to the Edelman Trust Barometer. The authors attribute this paradox to the multiple layers of reinterpretation in the current pipeline: different platforms deliver conflicting or oversimplified narratives, leaving audiences uncertain about what to believe and more likely to gravitate toward the most prominent or emotionally compelling voices.

The encouraging news is that trust itself is not the missing ingredient. Roughly 70 percent of Americans report at least moderate trust in scientists to act in the public’s best interest, and about 75 percent prioritize sources that demonstrate relevant medical expertise, disclose conflicts of interest, and present information clearly. Preferences vary across populations: people with lower incomes or less formal education are more likely to prioritize likability and relatability, while those without health insurance often emphasize convenience, including access through social media and AI platforms. The central weakness, the authors conclude, is a misalignment between trust, accessibility, and communication pathways. Their recommendations include having academic medical centers incentivize faculty and clinicians to engage in outreach and pair them with professional science communicators, as the Mayo Clinic has done by counting digital and social media scholarship among criteria for academic advancement. Community-level partnerships matter too: the Barbershop Talk with Brothers program, a collaboration between the Arthur Ashe Institute for Urban Health and SUNY Downstate Medical Center, trains barbers who are trusted advisors in Brooklyn to communicate evidence-based health information. Structural incentives, such as tying grants or accreditation to demonstrated public communication, and philanthropic support for programs like the ARVO Science Communication Training Fellowship could make outreach a core component of scientific careers. The challenge, in the authors’ framing, is not a deficit of public trust in science but the failure to ensure that accurate, transparent, and understandable information is easy to find, easy to engage with, and easy to trust, and philanthropy’s risk-tolerant capital may be the force best positioned to rebuild those connections.

Subject of Research: Redesigning the biomedical science communication ecosystem from publication to public engagement

Article Title: Bridging communication gaps across the biomedical ecosystem

Article References: Burke, E. A., Schumm, S. N., Jobe, A., Whelply, K., & Das, I. (2026). Bridging communication gaps across the biomedical ecosystem. Frontiers in Systems Biology, 6, Article 1939110. https://doi.org/10.3389/fsysb.2026.1939110

Image Credits: AI Generated

DOI: 10.3389/fsysb.2026.1939110

Keywords: science communication, biomedical research, peer review, misinformation, science journalism, social media, large language models, public trust, philanthropy, health information, news deserts, community health

The exchange centers on a cross-sectional study by Ismayilov and colleagues, published in September 2026, that surveyed individuals with cancer and their caregivers about their awareness of, attitudes toward, and experiences with generative artificial intelligence tools. The study appeared in Supportive Care in Cancer, a Springer journal devoted to the supportive dimensions of cancer treatment, and it arrived at a moment when large language models have moved from research curiosities to everyday consumer products. The original research team set out to map how this population, one of the most information-hungry in all of medicine, actually encounters and uses these tools in real life.

That mapping matters because the population in question is uniquely vulnerable to both the promise and the peril of generative AI. People living with cancer and the family members who care for them experience elevated rates of anxiety and depression across the illness trajectory, a pattern documented repeatedly in the psycho-oncology literature, including work showing that disease stage predicts post-diagnosis anxiety and depression only in some cancer types, and studies linking illness perceptions to psychological distress in lung and breast cancer patients. Distress drives information-seeking, and information-seeking now increasingly means conversing with a model that can produce fluent, confident, and occasionally wrong answers in seconds.

Into this landscape comes a correspondence from a team of Japanese nursing researchers led by Kohei Kajiwara of Shimonoseki City University, with Jun Kako of Mie University, Masamitsu Kobayashi of Toho University, Maki Kanaoka of Shimonoseki City University, and Makoto Yamanaka of Aichi Medical University. Their comment, published on 14 October 2026 as article 1098 in volume 34 of the journal, engages directly with the findings and the interpretation of the original cross-sectional study. All five authors are nurses or nurse scientists by training, several holding the oncology certified nurse specialist credential, a Japanese advanced practice qualification that signals deep clinical experience with cancer care. That clinical grounding shapes the perspective they bring to the debate.

The correspondence is more than a polite academic footnote. Comments of this kind are where the fine print of a study gets interrogated: how the sample was recruited, who might have been left out, what the survey questions could and could not capture, and how far the conclusions can reasonably travel. In this case, the commenters bring to bear their own body of prior work, including a rapid review published in JMA Journal in 2025 on artificial intelligence in nursing support for families. That review examined how AI technologies are being deployed to help family caregivers, a group the original study also placed at the center of its inquiry, and it forms part of the intellectual scaffolding behind the comment.

One of the most technically interesting threads running through the exchange concerns empathy. A systematic review by Sorin and colleagues published in the Journal of Medical Internet Research in 2024 examined whether large language models can be empathetic, compiling evidence on how these systems perform when users bring them emotional distress. The question is not trivial for cancer care. Caregivers of patients with head and neck cancer, for example, have been shown in prior research to perceive quality of life differently from the patients themselves, creating a gap in understanding that supportive care services try to bridge. If a chatbot offers a form of simulated empathy at three in the morning when no clinician is available, does that help, harm, or merely substitute for human connection? The reviewers found that the evidence base is still maturing, and the Japanese commenters lean on that uncertainty in their reading of the new survey data.

The technical architecture of generative AI explains much of this caution. Large language models generate responses by predicting likely text sequences based on patterns learned from vast training corpora. They do not consult a verified medical database, they do not know when their training data went stale, and they can produce outputs that are fluent and authoritative in tone while being factually incorrect, a phenomenon researchers call hallucination. For a patient weighing treatment options or interpreting a pathology report, the difference between a well-sourced answer and a confident fabrication can carry real clinical consequences. Cross-sectional surveys can reveal how often people reach for these tools and how they feel about them, but they cannot, by design, reveal how often the tools steer those people wrong.

There is also a structural problem with surveying awareness and attitudes in this population that the correspondence implicitly highlights. People who respond to surveys about AI are, almost by definition, people comfortable enough with the technology to have an opinion. Those with lower digital literacy, older patients, those without reliable internet access, and those whose illness leaves little energy for questionnaires are systematically underrepresented. Yet these are precisely the groups whose information needs in cancer care are often greatest. Any cross-sectional snapshot of AI use among patients and caregivers must be read with that selection effect in mind, and the commenters’ clinical experience with diverse patient populations gives them a vantage point on who the survey might have missed.

The timing of the exchange is significant in its own right. The original study was published on 7 September 2026, the comment was received on 11 September, accepted on 8 October, and published on 14 October, a turnaround of barely five weeks. That speed reflects how urgently the supportive care community wants to understand the collision between consumer AI and cancer care. The commenters declared no competing interests and received no external funding for their manuscript, and they reported that no datasets were generated or analyzed in the comment itself, marking it as a scholarly response rather than a new empirical study.

What emerges from the exchange is a picture of a field trying to keep pace with its own patients. The original survey by Ismayilov and colleagues provides a valuable baseline: it documents that individuals with cancer and their caregivers are aware of generative AI, that they hold measurable attitudes toward it, and that some of them are already using it in the course of living with the disease. The Japanese nursing team’s comment insists that this baseline be interpreted through the lens of clinical reality, where psychological distress shapes information behavior, where caregiver and patient perceptions diverge, where the empathetic capacities of language models remain scientifically unsettled, and where the evidence for AI supporting families is still being assembled. Neither document settles the question. Together, they sketch the research agenda that will: longitudinal studies tracking what patients actually ask chatbots, validation studies measuring how often the answers hold up against clinical evidence, and interventions that pair the accessibility of AI with the safeguards of professional oncology care. Until those studies arrive, the message from the nurses of Shimonoseki, Tsu, Ota, and Nagakute is a measured one: the enthusiasm is understandable, the adoption is already happening, but the evidence has not yet caught up with the enthusiasm, and in cancer care, that gap is exactly where patients can get hurt.

Subject of Research: Use of generative artificial intelligence tools by individuals with cancer and their caregivers

Article Title: Comment on: “The use of generative artificial intelligence tools among individuals with cancer and their caregivers: a cross‑sectional study of awareness, attitudes, and experiences”

Article References: Kajiwara, K., Kako, J., Kobayashi, M., Kanaoka, M., & Yamanaka, M. (2026). Comment on: “The use of generative artificial intelligence tools among individuals with cancer and their caregivers: a cross‑sectional study of awareness, attitudes, and experiences”. Supportive Care in Cancer, 34(11), Article 1098. https://doi.org/10.1007/s00520-026-11345-y

Image Credits: AI Generated

DOI: 10.1007/s00520-026-11345-y

Keywords: generative AI, cancer, caregivers, supportive care, large language models, oncology nursing, patient education, digital health, empathy, cross-sectional study, health information, artificial intelligence

The research team, led by Samar Thabet Jallad with Maysa Kassabry and Adel Taher Takruri, set out to answer a deceptively simple question: is there a measurable relationship between how nursing students perceive and accept AI and how confident they are in their own electronic health literacy? E-health literacy, a concept formalized in the widely used eHEALS instrument, describes a person’s perceived capacity to search for health information on the Internet, evaluate its credibility, and put it to practical use. In an era when patients arrive at clinics armed with search-engine results and clinicians rely on digital decision-support tools, that capacity has become a core professional competency rather than a bonus skill.

Methodologically, the study leaned on two validated instruments. The first was the eight-item eHealth Literacy Scale, which asks respondents to rate statements about their ability to access and appraise online health resources. The second was a 28-item questionnaire adapted from technology-acceptance research, organized into seven domains covering AI-related perceptions such as perceived usefulness, perceived ease of use, facilitating conditions, and anxiety, with negatively worded items reverse-scored where required. Together, the two scales produced a paired portrait of each student: one score capturing digital health confidence, the other capturing attitudes toward AI adoption in practice.

The headline numbers are striking for their balance. The mean eHEALS score came in at 3.53 on the scale used, with a standard deviation of 0.59, while the overall AI-related perceptions score averaged 3.35 with a standard deviation of 0.34. Both figures sit in the moderate-to-positive range, suggesting that these students neither worship nor fear the technologies arriving in their future workplaces. More revealing, however, was the statistical relationship between the two constructs. Overall AI-related perceptions correlated positively with eHEALS scores, with a Pearson correlation coefficient of 0.387 that remained significant after the researchers applied a Holm adjustment to guard against false positives across multiple tests.

Drilling down into the seven perception domains produced the study’s most technically interesting result. Facilitating conditions, a construct describing whether students believe the resources, infrastructure, and support exist to make AI usable, showed the largest domain-level correlation with e-health literacy at r = 0.450, again surviving Holm correction. In plain terms, students who felt that their environment would actually support AI use were also the ones most confident in navigating digital health information. Anxiety, by contrast, showed no significant association with eHEALS, a null result that challenges the common assumption that fear of new technology is the primary barrier to digital competence among health trainees.

To test whether the association was genuine rather than an artifact of demographics or Internet habits, the researchers built a multiple linear regression model. After adjusting for year of study, perceived Internet skills, and perceived importance of the Internet, overall AI-related perceptions remained independently associated with e-health literacy, with an unstandardized coefficient of 0.643 and a 95 percent confidence interval running from 0.289 to 0.998. The model explained roughly 17.7 percent of the variance in eHEALS scores, a modest but meaningful share for a psychological construct measured in a single sitting. Among the regrouped Internet-use comparisons, only perceived usefulness of the Internet remained significant after Holm correction, hinting that students who see the web as genuinely valuable are better positioned to build both AI acceptance and digital health literacy together.

The context of the study matters as much as its statistics. Palestine’s nursing education system operates under conditions that most Western digital-health research rarely confronts: intermittent infrastructure, restricted movement, and limited access to cutting-edge clinical technology. That 113 students in this environment report moderate e-health literacy and generally positive AI perceptions is itself a finding, suggesting that digital readiness among health trainees is not simply a function of national wealth or institutional budgets. The authors note that the relationship between AI perceptions and e-health literacy had been insufficiently examined in this setting, making the study a baseline against which future regional comparisons can be measured.

Yet the researchers are careful, almost insistently so, about what the data cannot show. The cross-sectional design captures a single moment in time, which means the correlation between AI attitudes and e-health literacy could run in either direction or reflect a third factor entirely. Students confident in their digital skills may simply feel more sanguine about AI, or positive exposure to AI concepts may build broader digital confidence, or both traits may grow from the same underlying comfort with technology. No causal claim survives the snapshot design, and the authors explicitly call for multi-site, longitudinal, and intervention studies to untangle the direction of the relationship.

There is also a measurement caveat that serious readers should weigh. Several of the adapted subscales demonstrated low internal consistency, meaning the domain-level findings, including the prominent role of facilitating conditions, should be interpreted cautiously. Internal consistency reflects how well items within a subscale hang together as a measure of a single construct, and when that cohesion is weak, domain-level correlations can be unstable. The overall AI-perceptions score and the eHEALS instrument are on firmer ground, but the granular story about which specific attitudes matter most will need replication with refined instruments before it can guide curriculum design with confidence.

Even with those limits, the study lands at a consequential moment. Nursing schools worldwide are racing to decide how much AI education to embed in their curricula, and evidence about which attitudes predict digital competence is scarce, particularly from low- and middle-income settings. If the association observed here holds up under longitudinal scrutiny, it would suggest that interventions targeting students’ perceptions of AI, especially their beliefs about whether supportive conditions exist, could pay dividends in general e-health literacy, a skill that directly affects patient education and evidence-based practice. For now, the Palestinian data offer a modest but well-adjusted signal: the students most ready to embrace intelligent tools are also the ones most confident in their ability to separate reliable health information from noise online, and that pairing may prove to be one of the most valuable competencies a twenty-first-century nurse can carry to the bedside.

Subject of Research: The association between artificial intelligence perceptions and e-health literacy among nursing students in Palestine

Article Title: Artificial intelligence-related perceptions and E-health literacy among nursing students in Palestine: A cross-sectional study

Article References: Jallad, S. T., Kassabry, M., & Takruri, A. T. (2026). Artificial intelligence-related perceptions and E-health literacy among nursing students in Palestine: A cross-sectional study. PLOS Digital Health, 5(9), e0001737. https://doi.org/10.1371/journal.pdig.0001737

Image Credits: AI Generated

DOI: 10.1371/journal.pdig.0001737

Keywords: artificial intelligence, e-health literacy, nursing students, Palestine, technology acceptance, eHEALS, cross-sectional study, digital health, nursing education, facilitating conditions, health information, PLOS Digital Health

Researchers SeolHwa Moon and Jeong-Eun Yang searched YouTube on May 6, 2023, using two Korean search terms for Alzheimer’s disease, logged out of the platform and sorted by the site’s default relevance ranking. From 364 videos initially identified, they applied strict eligibility criteria: the content had to concern Alzheimer’s disease, be uploaded between May 1, 2020 and May 1, 2023, and be in Korean. After screening titles, reviewing full videos, and removing duplicates and videos focused on other forms of dementia, 90 videos remained for detailed analysis. Each video was coded against a framework of 15 predefined content elements, twelve covering patient-centered topics such as awareness, causes, symptoms, protective and risk factors, screening and diagnosis, and treatment, and three covering caregiving-related information: challenges, available resources, and support.

The methodological rigor of the coding process was substantial. Two coders independently rated a subset of videos before and after the main coding, achieving Cohen’s kappa values between .777 and .834 for categorical variables and intraclass correlation coefficients between .873 and .924 for continuous measures, levels the authors describe as substantial to almost perfect agreement. Quality was assessed with the Global Quality Score, a five-point scale in which scores of 1 to 3 indicate poor quality and scores of 4 to 5 indicate good quality, while reliability was measured with a five-item subset of the DISCERN instrument, which probes whether information goals are clear, sources are credible, presentation is balanced, additional references are listed, and areas of uncertainty are acknowledged.

The headline finding on quality is broadly encouraging. The mean Global Quality Score was 3.29, and exactly half of the videos, 45 of 90, were rated as good or excellent, with 30 percent achieving the top rating. The mean DISCERN reliability score of 3.58 out of 5 also pointed toward generally trustworthy content. This contrasts favorably with a broader systematic review of health-related YouTube content, which reported a mean Global Quality Score of just 2.68 and a mean DISCERN score of 2.36. The authors suggest that national policy efforts in South Korea, including the national dementia responsibility system implemented in 2017, may have helped raise the caliber of publicly available dementia information in the country.

Who uploads the videos matters enormously. Healthcare institutions and health channels accounted for 58.9 percent of the analyzed videos, individual users for 24.4 percent, and broadcasters for 16.7 percent. When the researchers split the sample into good-quality and poor-quality groups of 45 videos each, the differences became statistically sharp. Healthcare institutions or health channels produced 71.1 percent of the good-quality videos but only 46.7 percent of the poor-quality ones, a difference the authors report as significant with a Cramér’s V of 0.34. Healthcare professionals appeared as speakers in 78 percent of good-quality videos versus 49 percent of poor-quality ones, while journalists and interviewers were far more common behind the microphone in the poor-quality group. In short, the presence of a credentialed medical voice was one of the strongest markers of content quality.

Video length told a similar story. Good-quality videos averaged 19.57 minutes, nearly triple the 7.48-minute average of the poor-quality group, a difference with a large effect size of Cohen’s d equal to 0.81. Longer videos apparently had the room to explain symptoms, causes, and diagnostic procedures in depth, whereas shorter clips tended to deliver fragmentary information. Interestingly, the number of views, daily views, and comments did not differ significantly between quality groups, though good-quality videos did receive significantly more likes. Popularity, in other words, is not a reliable proxy for accuracy, a point with real consequences for viewers who assume that trending content is trustworthy content.

The most troubling finding concerns what the videos leave out. Good-quality videos most frequently covered symptoms, appearing in 93.3 percent of them, causes, in 91.1 percent, and screening or diagnosis, in 77.8 percent. But caregiving experiences appeared in only 4.4 percent of good-quality videos, information about available resources in 11 percent, and support in just 2.2 percent. Poor-quality videos were even thinner, with caregiving challenges and resources each mentioned in 2.2 percent of cases. On average, good-quality videos addressed 5.93 of the 15 content elements, while poor-quality videos covered only 3.20. The online conversation about Alzheimer’s disease, it turns out, is overwhelmingly a conversation about the disease itself and almost never about the daily reality of living with it or caring for someone who does.

This gap matters because the evidence on caregiver needs is unambiguous. Previous research reviewed by the authors found that non-professional caregivers want information not only about the disease but also about providing daily patient care, accessing health services, and looking after their own wellbeing. Caring for a person with Alzheimer’s disease requires knowledge about monitoring early symptoms, navigating diagnosis, assisting with instrumental activities of daily living, and preparing for possible long-term care placement. Surveys indicate that around 70 percent of caregivers prefer the internet and electronic media when collecting care information, which makes YouTube’s silence on practical care a genuine public health problem rather than a mere content gap. Better health literacy among caregivers has been linked to reduced caregiver burden and improved caregiving outcomes, so the platform’s blind spot translates directly into unmet need.

There is also a visibility problem at the top of the search results. Poor to generally poor-quality videos made up 38.9 percent of the sample, and although only a few of them promoted treatments or diets without scientific references, hyped new drugs, or selectively presented conference findings, some of these videos ranked at the top of search results. That placement increases their accessibility to patients and caregivers and could steer vulnerable viewers toward incorrect medical decisions. The authors note that visual and technical factors they did not measure, such as thumbnails, titles, and search engine optimization, can attract viewers through exaggerated or misleading representations regardless of actual content quality, compounding the risk.

The study’s conclusions come with honest caveats. The analysis covered only Korean-language YouTube content at a single point in time, so results may not generalize to other platforms, countries, or languages, and YouTube’s search algorithms and personalization may have shaped the retrieved sample. Most high-quality videos came from medical institutions, which may have biased the sample toward expert-centered perspectives and underrepresented patients, family caregivers, and advocacy groups. Nonetheless, the overall picture is clear and actionable: Korean-language YouTube videos about Alzheimer’s disease are generally of high quality and reliability, but they concentrate on disease knowledge while leaving practical caregiving information scarce. The authors call for standardized tools to help consumers judge video credibility, and urge that future content be developed with experts such as registered nurses, social workers, and care providers, ideally in collaboration with caregivers themselves, so that the platform that families already trust can finally answer the questions they are actually asking.

Subject of Research: Quality and content analysis of Korean-language YouTube videos about Alzheimer's disease

Article Title: What about Alzheimer 's disease online information?: YouTube video content analysis

Article References: Moon, S., & Yang, J.-E. (2026). What about Alzheimer 's disease online information?: YouTube video content analysis. Heliyon, 12(15), Article e45544. https://doi.org/10.1016/j.heliyon.2026.e45544

Image Credits: AI Generated

DOI: 10.1016/j.heliyon.2026.e45544

Keywords: Alzheimer's disease, YouTube, health information, caregivers, dementia, video content analysis, Global Quality Score, DISCERN, health literacy, South Korea, online health information, patient education