By Michael J. Critelli | MakeUsWell Newsletter, 

We are living in a world in which nutrition guidance has never been more available or more dangerous.

For years, social media has been filled with exaggerated claims about foods, beverages, vitamins, and supplements that supposedly “cure” difficult medical conditions. Artificial intelligence now makes those claims look and sound more believable. Deepfake videos, once difficult and expensive to create, can now be produced quickly, cheaply, and convincingly by many more people.

People dealing with frightening or hard-to-treat conditions are understandably vulnerable. When someone is in pain, losing sleep, frightened about cognitive decline, struggling with a chronic condition, or frustrated by conventional medical care, they may try almost anything. A persuasive video featuring a trusted celebrity, physician, or news anchor can become a powerful sales tool for a questionable remedy.

Because of the huge interest in weight loss solutions, one of the most widely watched deep fakes appeared to have Oprah Winfrey endorsing a morning wellness routine called the “pink salt trick,” a  mixture of water, Himalayan pink salt, and lemon juice which someone drinks on an empty stomach. In this video, the fake Oprah  “interviews” Dr. Rachel Goldman, whom she acknowledges to be the inventor of this routine. Nothing is authentic in this video, not Oprah or Dr. Goldman, and not their remarks.

As someone interested in therapies for tinnitus, I had a similar experience last year.  I saw a video that appeared to show Dr. Dean Ornish, the renowned cardiologist, promoting a supplement for tinnitus. Because I know Dr. Ornish, I paid close attention. Within a few minutes, I realized that the video was a deepfake.

That experience reinforces why we are building our food intelligence product with great care, including the way we explain its purpose, limits, and appropriate use.

Disclosures matter. But what I have learned over a lifetime is that disclosures overloaded with every imaginable risk often protect the drafter better than they educate the user. 

We want a different standard: communications that are clear, readable, and useful at the moment a user needs them. Our goal is an end-to-end process in which all of our communications educate users by preventing and clearing up misunderstandings.

Nutrition and health guidance are highly fact-specific. Medical researchers and nutrition experts often disagree. Sometimes the evidence is strong. Sometimes it is preliminary. Findings may be inaccurate, obsolete, overstated, or simply inapplicable to a particular person. 

One core feature of our product is an AI-supported process that not only searches, but validates and fact-checks information used in response to a user’s request. 

We also, of course, encourage users to work collaboratively with licensed health professionals who know their health history, medications, lab results, risks, and personal circumstances. That professional check-and-balance is essential and it is also a tenet baked into our product, its foundational design principles, and how it interacts with users.

The interplay between nutrition and medical conditions can be quite complex. I have tinnitus, which for me means hearing a persistent humming sound. YouTube has hundreds, if not thousands, of videos promoting vitamins, supplements, foods, and devices as tinnitus “cures.”

But the word “tinnitus” itself can be misleading. It is not a single disease with one cause and one cure. It is an umbrella term for a symptom that may have multiple origins. Some forms of tinnitus may be treatable. Others may not be. Correcting a Vitamin B-12 deficiency, for example, might matter in a subset of cases. But that does not mean that or any other vitamin supplement cures tinnitus.

Whenever someone claims that a complex, multi-factorial condition with uncertain origins can be cured by a single food, vitamin, or supplement, we should be immediately skeptical.

This skepticism is central to our product design. We do not want simply to give users answers. We want to help them understand the quality of the evidence behind those answers. Is the research strong or weak? Is it contested? Does it apply broadly, or only to a narrow population? Is it based on a clinical trial, an observational study, a mechanistic theory, or marketing language dressed up as science?

This is where AI can be extraordinarily valuable. It can help people ask better questions, compare competing claims, identify red flags, summarize research, and become more informed consumers and managers of their own health.

But the same technology can also be used to mislead them.

That is why food intelligence cannot be casual, simplistic, or overconfident. It must be careful, transparent, evidence-aware, and humble. It must help users proceed thoughtfully, monitor how they respond to meaningful dietary or supplement changes, and involve licensed professionals when the issue touches diagnosis, treatment, medication, supplementation, or a serious medical condition.

AI is a tool. Used well, it can empower people. Used badly, it can exploit their fear and hope.

Our responsibility is to build the kind of food intelligence product that helps people tell the difference.