How can consumers identify deepfake videos in online health ads?
Executive summary
Consumers are being targeted by increasingly convincing AI-generated videos that impersonate doctors and public-health figures to sell unproven treatments and supplements, and spotting them requires a mix of visual skepticism, source verification and technical help because platforms and automated detectors still struggle to keep pace [1] [2] [3]. Practical checks — scrutinizing the ad’s provenance, audio-visual glitches, dubious endorsements and inconsistent supporting claims — plus using platform reporting tools and reputable verification resources can reduce risk, although no single method is foolproof [4] [5] [3].
1. Look for the provenance: who uploaded it, where and why
A strong first clue is the ad’s origin: many deepfake medical ads appear on social platforms and are linked to accounts or pages that primarily sell supplements or have short lifespans; investigative reporting has repeatedly shown networks using cloned clinician footage to promote products across TikTok, Instagram and other sites [6] [7]. If the video appears as a sponsored ad or is shared from an unfamiliar account with minimal history, that’s a red flag — real clinicians and institutions rarely run high-volume paid campaigns through anonymous merchant accounts [2] [8].
2. Read beyond the talking head: verify claims and citations
Deepfake “doctors” often make sweeping claims or present fabricated endorsements like fake “FDA certificates” or miracle testimonials; TODAY’s investigation found such fabricated artifacts used to sell weight-loss and diabetes products [1]. Cross-check any clinical claim by searching for the researcher, institution or study cited; reputable trials and regulatory approvals will appear on academic or government sites, while bogus product claims won’t be corroborated by PubMed or public-health organisations [9] [10].
3. Watch closely for telltale audio‑visual artifacts and contextual inconsistencies
Even sophisticated deepfakes can betray themselves in small ways: lip-sync drift, unnatural blinking or micro‑expressions, mismatched lighting between face and background, robotic prosody or oddly clipped breaths in the audio; multiple outlets and technical reviews note that these “slop” artifacts remain common in synthetic medical videos [3] [5]. Also look for mismatches in attire, backdrop or references — a clinician claiming a current role but shown in footage from an old talk is suspicious and has been used in documented cases of impersonation [2] [6].
4. Use platform tools and third‑party resources but don’t assume they’re perfect
Platforms offer reporting and, sometimes, labels for manipulated media; Australia’s eSafety Commissioner and fact‑checking groups provide guides to spot deepfakes and report abuse [4] [6]. However, enforcement is uneven: cases have persisted online for weeks before removal, and automated detection still lags attackers, so reporting suspected deepfakes and checking fact‑check organizations (Full Fact, investigative journalism outlets) remain essential [2] [7] [5].
5. Simple verification steps that work fast
Pause before clicking: reverse-image or reverse-video keyframes to find the original source; search the doctor’s name plus “deepfake” or “statement” and check the purported institution’s official site or social channels for confirmations or denials — Diabetes Victoria and the Baker Institute publicly disavowed deepfaked endorsements in a widely viewed case [4] [9]. If a video urges immediate purchase, uses high-pressure language, or offers “certificates” without links to verifiable regulators, treat it as likely fraudulent [1] [8].
6. Know the limits and protect the vulnerable
Detection tools and literacy campaigns help, but they do not eliminate the threat: reviews of detection research show continual cat-and-mouse dynamics between generation and forensic methods, and elderly or low-health-literacy consumers are especially at risk of persuasion by faux experts [3] [11]. When in doubt, seek a trusted clinician’s advice before acting on a health ad, and encourage platforms and regulators to require transparent AI‑labelling and faster takedowns — policy responses are being proposed in multiple jurisdictions because technical fixes alone won’t solve the problem [12] [8].