How AI Stealth Influencers Are Becoming Health Influencers
Look better and feel better – Those advertising claims are irresistible, especially for those of us with aging, imperfect bodies, which is why there’s such an appetite for products that boldly promise to “Clear your skin,” “Stop the pain,” “Give you more energy,” and even “Help you live longer.” Our bodies are vital, so why are people increasingly taking product advice from endorsers who don't really have one?
The “no-bodies” are AI marketing influencers, which have already been around for many years – I can hardly believe it’s been seven years since I first wrote about “Artificial Endorsers” and described Miquela Sousa, or Lil Miquela, who was likely the best-known virtual spokesperson at the time.
When I penned that piece in July of 2019, AI endorsers were novelties and even to the untrained eye, fairly obvious fakes. In August 2026, virtual influencers are much more common and much harder to detect, which should make us ask: In another seven months, let alone seven years, will we be able to tell if a spokesperson is human or AI?
The answer may not matter to some because human-like AI is already pervasive and largely accepted, e.g., customer service phone support, grocery store self-checkouts, online chat, smartphone virtual assistants. The thinking me be, ‘What’s a few more?’ or, “Why worry if the influencer endorsing a Prada handbag is Lil Miquela/AI or Hunter Schafer/human?’
However, there are several reasons why it should matter if spokespeople is human or AI. In a May 2023 article, “The Pros and Cons of Artificial Endorsers,” I identified the following four strikes against artificial endorsers:
Loss of jobs for humans
Lack of authenticity by bots
No lived experience for AI
AI’s utter bias
Three years later, the analyses still seem relevant. However, what’s since changed dramatically is that extremely realistic-looking AI endorsers, whose bodies are 100% digital, are increasingly used to promote products that directly affect flesh and blood, i.e., humans’ health.
One of the stealth AI wellness endorsers is Melanskia, a plainly dressed, bonnet-wearing older woman whose appearance and Instagram video surroundings suggest she’s Amish. Of course, she can’t really be Amish because she’s purely digital. Her human creator also appears far-removed from anything Amish: He’s Josemaria Silvestrini, a 28-year-old entrepreneur who lives in Shanghai where he runs the company whose product Melanskia persistently promotes.
In her wide array of Instagram reels, Melanskia confidently imparts ‘Amish-inspired,’ health-related insights through intriguing topics such as:
Five spices forbidden in my Amish community
The most dangerous item in your kitchen
How to spot the difference between real and fake butter
Never pour hot food in plastic
Much of the advice she gives seems plausible, like being wary of nonstick coating peeling off pans and into food. However, hearing her counsel, it’s reasonable to wonder how someone who in the real world would likely have an 8th grade education became an expert in toxicology, pharmacology, and biochemistry. Unfortunately, for those not sharing such cynicism, “There is no disclosure that everything about [Melanskia] is made by A.I. tools.”
It’s also ironic for an Amish person, but not surprising for an influencer in this age of hyper-commercialization, that each health-related problem Melanskia identifies leads to a singular solution – Modern Antidote, a product available on Amazon.
Modern Antidote’s positioning as a “support drink mix” places it in the largely unregulated realm of dietary supplements. The National Institutes of Health describes the lower regulatory threshold for supplements versus medicines:
“Medicines must be approved by FDA before they can be sold or marketed. Supplements do not require this approval. Supplement companies are responsible for having evidence that their products are safe, and the label claims are truthful and not misleading. However, as long as the product does not contain a new dietary ingredient (one introduced since October 15, 1994), the company does not have to provide this safety evidence to FDA before the product is marketed.”
So, unlike medicines that require exacting clinical trials and painstaking applications to secure FDA approval, supplements make it to market much more easily based simply on loose evidence that they are safe and that the claims are “truthful and not misleading.”
For pharmaceuticals, accountability for recommendations rests squarely on prescribing physicians beholden to the Hippocratic oath, whereas the integrity of communication about dietary supplements depends largely on the virtues of the organizations that make them and the individuals who promote them, which brings us back to AI influencers and Melanskia.
In August 2022, amid the proliferation of social media influencers, I wrote, “Are there Rules When Everyone’s an Endorser?” Although this article focused on human spokespeople, its four key rules apply to virtual influencers too:
Know the product
Believe in the product
Ensure the product is a good fit for the target market
Disclose your relationship with the organization
Generally speaking, AI excels in #1 knowing products – the amounts of relevant information it can distill is staggering. It’s also adept at asking questions to clarify needs and find optimal solutions – #3.
However, those capacities apply to general chatbots like Claude and ChatGPT, which have no obvious product biases. Their objectivity is in stark contrast to that of AI endorsers, which are strictly controlled by the companies that create them so that their digital agents to do/say whatever their makers dictate.
Human spokespeople also tow the company line, or communicate the corporate message, sometimes to a fault, but there are several key differences with moral implications that apply to virtual influencers:
Lack of Volition: AI agents are 100% compelled to do their principals’ bidding. Unlike human spokespeople, virtual influencers can’t decide whether they want to rep a particular product or say a specific statement. AI endorsers are made to do whatever their creators want.
No Conscience: The lack of free will becomes all the more problematic because virtual influencers also are unlikely to have a moral compass. So, even if they had volition, if a product they’re promoting holds potential for harm or deception, they won’t recognize the ethical issue.
Lack of Lived Experience: Recommendations tend to carry more weight when they come from people who have used the products in question. AI endorsers are very good at acting like they’re users and even saying that they do use the products they promote, but they’re completely digital DNA makes product use for them impossible.
No Other Clues: Sometimes we can tell by the expression on a human endorser’s face or by their body language that they don’t truly believe in the product. We also gather meaning from things they do and say outside the advertising, e.g., When LeBron James, who was a McDonald’s spokesperson at the time, famously told reporters that a key to his staying in playing shape was not eating at McDonald’s. Virtual influencers don’t exist outside their ads, so we can’t see if they practice what they preach, and in the ads, their words and actions can be carefully controlled to avoid giving any hint of inauthenticity.
These differences can have negative impacts even when we know an endorser isn’t human. For instance, just watching Melanskia for a few minutes, it’s easy to forget she’s AI and start to think that she really does make her own butter, cook in cast-iron pans, and benefit by using Modern Antidote.
When consumers don’t realize that influencers are virtual, dangers from these four discrepancies expand and are continually active. The potential for harm also increases when the products are items that people ingest, such as dietary supplements like Modern Antidote.
