Large language models are now among the first go-to sources for health information. Yet little is known about where these systems actually get their facts.
A new report titled “AI’s Sources of Truth: How Chatbots Cite Health Information” reveals where major language models like ChatGPT, Gemini, Claude, and Perplexity get their medical advice.
Researchers from Outcome Rockets examined over 5,400 website links shared by these chatbots to determine which sources they trust most when answering health-related questions.
They Like the Easy Stuff
The study found that chatbots prefer content that’s simple and easy to understand. Nearly one-third of all links came from health media sites such as the Mayo Clinic and the Cleveland Clinic.
These trusted sources, along with the government’s PubMed Central, top the list. This shows that AI tools favour info reviewed by medical experts but still written for everyday readers.
What’s surprising is how little they rely on academic research. Only about 23% of their citations came from direct studies. Instead, they mostly relied on summaries and blogs.
About 59% of their sources are simplified versions of real studies. The type you will find in blog posts. They also often use info from business blogs or company pages.
Trustworthy and Up-to-Date
Around 62% of the links chatbots share come from trusted and familiar websites like the National Institutes of Health. This means that most of the info they dish out to users is credible.
Also, about 67% of the articles cited by these chatbots were from either 2024 or 2025. This shows that the data is fresh and not outdated. Only 0.2% of the links were broken, meaning that users rarely hit a dead end.
How the Bots Differ
Each chatbot has its own way of finding sources. Gemini focuses more on government and nonprofit health pages, making it the most policy-oriented.
ChatGPT relies a lot on health media sources and is the one most likely to share a link that requires payment to read.
Claude is more balanced, using a mix of health news and research studies in its answers.
Perplexity tends to use more commercial websites and user-generated content like YouTube.
Even with their differences, almost all the links shared by these chatbots, about 99%, are free for anyone to open. Most answers include around a dozen sources, giving users plenty to explore.
The study shows that AI chatbots are changing how people look for health info. They rely on clear and reliable sources, but it is still smart to check their answers.
Source from Gizchina
Disclaimer: The information set forth above is provided by gizchina.com independently of Alibaba.com. Alibaba.com makes no representation and warranties as to the quality and reliability of the seller and products. Alibaba.com expressly disclaims any liability for breaches pertaining to the copyright of content.
