A systematic review published on January 29, 2026, found that generative AI increased the volume, speed and perceived credibility of health misinformation in the available research. The BMC Public Health review included 15 empirical studies published between January 2023 and August 2025 and found that people often struggled to distinguish AI-generated health misinformation from human-written material.

The finding complicates the case for clinicians using AI to produce public-facing posts, videos and graphics. The tools can turn source material into a draft quickly, but a professional remains responsible for checking what the draft says, what it leaves out and whether the evidence applies to the audience described.

Health professionals have a role beyond the clinic

Social platforms allow health professionals to explain evidence directly to a large audience, without waiting for a consultation or a traditional media interview. The World Health Organization (WHO) has treated that channel as part of public-health communication. A WHO-hosted 2021 toolkit described doctors and nurses as effective, trusted messengers and set out ways for them to create, reuse or share reliable vaccine information during the COVID-19 pandemic.

That toolkit addressed a specific emergency and vaccine campaign, not every health topic or platform. Its premise remains useful: a medical qualification can give a post authority, but the published claim still needs a traceable source and an honest account of uncertainty.

A broader evidence review shows why the task is difficult. A 2024 Journal of Medical Internet Research review synthesized 70 English-language sources published from January 2012 through March 2024 and found broad agreement that health misinformation on social platforms is a problem. The authors said health professionals were well placed to respond, but workload and stress limited their ability to do so.

Diagram showing three goals of reliable health communication: reducing pseudoscience, improving health literacy and building trust

AI changes production time, not the evidence threshold

Generative AI can draft text, summarize documents and propose visual or video structures. That makes it useful for moving from a blank page to material a clinician can edit. It does not show whether a paper is current, whether a numerical result applies outside the study population or whether a statement has crossed from education into an unsupported treatment claim.

WHO's 2024 guidance on large multi-modal models says health-focused systems can produce false, inaccurate, biased or incomplete statements and may encourage automation bias. The guidance calls for defined tasks, stakeholder involvement, post-release auditing and independent assessment. Those controls matter before publication because fluent wording can conceal a missing source or an overstated conclusion.

A workable editorial process separates machine assistance from professional approval. AI may organize questions, compare a draft with supplied sources or convert an approved outline into several formats. The human publisher still needs to open the original source, confirm every number and date, preserve limits stated by the researchers and remove any citation that cannot be found.

Short posts still need context

Platform formats reward compression, while medical evidence often depends on population, comparator, outcome and time period. Removing those conditions can turn an accurate study result into a misleading general claim. A post should therefore identify the institution or paper behind a claim and state whether the finding comes from a trial, an observational study, a review or public guidance.

Updates also need a process. Guidance changes, studies are corrected and regulatory decisions differ across countries. A creator who cannot revise an old graphic or attach a correction risks leaving a widely shared claim online after its evidence has changed.

The 2026 generative-AI review also limits how far its own conclusions can travel. Its authors called for more real-world testing across languages, platforms and formats, especially audio and video. The available evidence supports caution about scale and credibility; it does not measure the accuracy of every model or every clinician-led account.

Professional authority can magnify commercial claims

Clinical credentials may make product endorsements or treatment claims more persuasive. The World Medical Association's October 2024 guidance says physicians appearing in mass media should provide objective, evidence-based information, avoid unproven procedures or products and include material risks when discussing interventions. It also says physicians should not use media appearances to recommend named products or exaggerate their qualifications for commercial benefit.

A clinician reviews a health communication draft and its publication requirements (illustrative image)

That professional guidance is international, but it is not a substitute for national law. Rules governing health advertising, patient confidentiality, professional titles, sponsorship disclosures and remote interactions differ by country. The source article cited Taiwan's rules for medical institutions and personnel; those rules cannot be presented as a global legal standard.

Trust depends on a visible verification chain

Follower counts and posting frequency do not measure accuracy. Readers can inspect a stronger set of signals: whether claims link to identifiable evidence, whether credentials and financial relationships are disclosed, whether uncertainty is stated and whether corrections remain visible.

For clinicians, AI can reduce the labor of producing a draft, but it also makes unsupported material easier to publish at scale. The defensible division of work is narrow: machines can assist with production, while people remain accountable for evidence, context, privacy and the final decision to publish.