Cheap Automated Ethos: AI healthcare scams
FBI and FTC data have recorded the rise in scams and fraud in recent years. One of the most affected fields has been healthcare which is particularly vulnerable to fraud due to its high stakes. Whereas in the past ethos--or credibility--used to be hard to fake, now deepfakes can imitate medical professionals for cheap and professional websites can be created easily using LLM tools. Old signals for detecting fraud are being made obsolete.
COMMENTARYBEGINNER FRIENDLY
Sohum Bhardwaj
2/25/20263 min read
Cheap and Automated
Ethos is a term that refers to credibility and trust. In healthcare, having a medical degree gives you ethos; it shows that you spent years studying medicine and human health at an institution that held you accountable. It is not a stretch to say that we rely on ethos often. Most people don't question their doctor's expertise or a medicine's legitimacy. When a website looks professional, has a large number of reviews, and we can see youtubers raving about how great the product is, the product earns our trust. And reasonably so, most people don't want to dig through research and spend hours verifying medical claims.
Essentially, ethos is a shortcut. It spares us from expending effort and constantly questioning everything around us. It is reliable when it is hard to fake, but coding agents, LLMs, and video generation models are changing the bar for quality faster than we can adapt.
What happens when anyone can generate a professional looking website, believable customer reviews, and a deepfake video of a doctor who claims that their medicine will cure your glaucoma? We used to be able to rely on quality as a proxy for ethos, but quality is increasingly becoming cheaper to imitate.
This image is an example of AI generated ethos. The certificates are fake. Sourced from this news article: Woman Scammed by Online Ad With Deepfake of Her Doctor Speaks Out
The FBI internal crime report for 2025 shows a clear trend of increasing scams and crime. In 2025, they have started to track AI-related scams and fraud as well as of 2025, and the numbers are staggering. While AI is only used in a small portion of all crime tracked by the FBI, it cost Americans nearly $900 million out of around $20 billion in total losses (Home Page - Internet Crime Complaint Center (IC3))
Another stunning jump the amount of money lost to healthcare fraud which increased from $20.7 million to $84 million between 2023 and 2024 according to data from the FTC Consumer Sentinel Network (FTC). I skimmed the FTC consumer network annual report, but I couldn't find the reason for this drastic change.
This image shows the rise in scams and fraud that continues to increase despite increased education. Sourced from the IC3 annual report: Home Page - Internet Crime Complaint Center (IC3)
The rising costs of healthcare scams and overall fraud as seen in the FTC and FBI's data illustrate the increasing severity of fraud on the US economy. AI has the potential to accelerate this trend, allowing scammers to widen their reach and craft more detailed and convincing schemes. The truth is we need to adjust to a new era where surface-level quality no longer indicates ethos.
Other measures of ethos like third-party reviewers need to become the default rather than trusting reviews on a product's own website. These third-party websites and reviewers already exist, but they can be vulnerable to fake reviews and spam. In the future, they will be even more important.
We also need to reform what is taught in digital literacy to include AI misinformation. Currently, students are taught about misspellings and incorrect formatting; these signals will fade away due to AI. Things like urgency and exaggerated claims won't fade away since they are core parts of scams. And new tactics such as the use of AI checking platforms should be taught.
While ethos won't disappear due to AI, it will definitely change. In the future, we will likely use completely different metrics to evaluate the believability and trustworthiness of a piece of media.