Blog
When AI Says “Unproven”: Are We Researching Health Widely Enough?
When AI Says “Unproven”: Are We Researching Health Widely Enough?
Artificial intelligence has changed the way we search for answers.
A question that once required hours of reading can now produce a detailed explanation within seconds. In healthcare, this can be extremely useful. AI can summarise research, explain difficult concepts and help people explore subjects that were once much harder to understand.
But easier access to information brings another question.
What happens when an answer is reasonable, well written and supported by credible sources, but still does not represent the whole picture?
The problem may not necessarily be misinformation.
Sometimes, we simply have not looked widely enough.
A Correct Answer Can Still Leave the Wrong Impression
When we ask AI a health question, it naturally gives considerable weight to established medical institutions, systematic reviews, major scientific publications and widely recognised standards of evidence.
That is generally a sensible place to begin.
OpenAI itself cautions that ChatGPT can sometimes oversimplify, present a single perspective as absolute truth, or express confidence in an incorrect answer. OpenAI: Accuracy and reliability
However, the most visible information is not necessarily the entirety of available knowledge.
Research may be concentrated in particular countries. Important regulatory history may exist in another language. Clinical experience may develop for decades before researchers become interested in studying its mechanism. Traditional observations may exist long before anyone designs a modern clinical trial around them.
Different countries can also study, regulate and use the same therapy in very different ways.
This creates an important distinction.
An AI system may conclude that there is currently insufficient high quality evidence to establish a particular treatment conclusively.
A reader may interpret that as meaning there is no credible basis for the treatment at all.
Those are not the same conclusion.
“Not Proven” Does Not Mean “Proven Not to Work”
Scientific knowledge develops gradually.
Many discoveries begin with observation. Someone notices an effect. Others notice something similar. Questions are raised. Possible explanations are proposed. Research follows and, if the subject continues to show promise, better studies are eventually conducted.
Sometimes the original observation survives this process.
Sometimes the explanation changes completely.
Sometimes the effect turns out to be smaller than originally believed.
Sometimes it disappears when properly tested.
That is exactly what scientific investigation is supposed to determine.
At any point in this process, however, saying that we do not yet have enough evidence is different from saying that something has been demonstrated not to work.
The first statement describes the present state of knowledge.
The second makes a conclusion about the treatment itself.
The difference may appear small, but it can completely change how people approach an unfamiliar field.
Traditional Chinese Medicine Shows Why the Question Matters
Consider a simple question such as:
Does Traditional Chinese Medicine work?
Even the question is difficult because Traditional Chinese Medicine is not one treatment.
It includes herbal medicine, acupuncture, moxibustion, therapeutic movement, dietary practices, traditional diagnostic methods and a very large number of individual formulations and interventions.
Trying to assign one evidence rating to all of Traditional Chinese Medicine would be similar to asking whether Western medicine works.
Surgery, antibiotics, physiotherapy, chemotherapy, nutritional therapy and antidepressants obviously cannot be evaluated as a single intervention.
Traditional medicine deserves the same precision.
Some practices may eventually demonstrate clear clinical value. Some may offer modest or situation specific benefits. Some traditional explanations may describe genuine observations even if modern science later explains those observations through a different mechanism.
Others may remain uncertain.
Some may eventually be shown to have little or no meaningful effect.
Taking traditional knowledge seriously does not mean accepting everything within it.
It means being willing to investigate individual practices fairly.
Experience Is Not Scientific Proof, but It Is Not Meaningless
Modern medicine places great importance on controlled research for good reason.
Personal experience can be influenced by expectation, coincidence, natural recovery and many other factors. One person's experience therefore cannot establish that a treatment will have the same effect on everyone.
However, observation still has a place in the development of knowledge.
Medicine has repeatedly advanced because somebody noticed something interesting and decided that it deserved further investigation.
Traditional medical systems contain generations of accumulated observations. Some may ultimately prove valuable. Some may not.
The role of science is to help us distinguish between them.
This is why personal experience and traditional observation should not automatically be treated as proof, but neither should they automatically be treated as worthless.
They can be the beginning of a question.
The International Conversation Is Changing
The global discussion around traditional and complementary medicine is also becoming more sophisticated.
In 2025, World Health Organization Member States adopted the WHO Global Traditional Medicine Strategy 2025 to 2034.
The strategy calls for strengthening the evidence base for traditional, complementary and integrative medicine, improving safety and regulation, and integrating safe and effective approaches into health systems where appropriate.
That position is important because it avoids two extremes.
It does not say that a treatment works simply because it is traditional.
It also does not say that traditional knowledge should be discarded because every part of it has not yet been validated by modern research.
Instead, the approach is to investigate it, improve the evidence, regulate it appropriately and determine what can genuinely contribute to healthcare.
This may also be a useful model for the way artificial intelligence handles these subjects.
AI Can Inherit the Narrowness of the Information It Sees
Artificial intelligence can access and process an extraordinary amount of information.
But enormous does not mean complete.
Much of the health information most visible online is written in English and shaped by major English speaking medical institutions, publications and research databases.
These are valuable sources, but they do not represent the whole of human medical knowledge.
Important research, regulatory history and clinical experience may exist in Japanese, Chinese, Korean, Hindi, Arabic, Persian, Russian, Spanish, Portuguese and many other languages.
Useful information may come from regional universities, hospitals, government agencies, medical journals, professional communities or traditional medical systems that receive relatively little attention in mainstream English language discussion.
Sometimes the information is not absent.
It is simply harder to find, less frequently translated, less widely cited, or contained within a different academic or regulatory system.
This matters because the way we conduct research can influence the conclusion we reach.
If an investigation begins and ends with a small group of familiar sources, the answer may be accurate within that particular frame while still missing meaningful knowledge from elsewhere.
AI can therefore create an unusual problem: confidence without completeness.
There Are Different Kinds of Evidence
Another source of confusion is the tendency to treat all forms of evidence as though they answer the same question.
They do not.
Historical use tells us that people have repeatedly chosen to use something.
Clinical experience may generate observations worth investigating.
Laboratory research may help identify a possible biological mechanism.
Clinical trials attempt to determine whether an intervention produces meaningful results in people.
Systematic reviews examine the strength and consistency of those studies.
Regulatory recognition tells us something about how an authority has classified, evaluated or controlled a particular type of therapy or device.
None of these automatically proves everything claimed about a treatment.
A regulated medical device is not proof of every possible therapeutic claim.
A positive clinical trial does not establish that every variation of an intervention works.
Centuries of use do not guarantee effectiveness.
Personal experience does not establish universal efficacy.
But neither should these different pieces of information be ignored simply because none of them can answer every question on its own.
Good research requires understanding what each piece of evidence tells us and where its limits are.
The Answer Is Not Lower Standards
A broader approach to traditional and complementary health should never mean abandoning scientific standards.
Traditional treatments can be ineffective.
Herbal products can interact with medications.
Poor quality products can contain contaminants.
Claims can be exaggerated.
Commercial interests exist in every part of healthcare.
Safety, quality control, appropriate regulation and good research remain essential.
The answer is therefore not to lower the standard of evidence.
The answer is to widen the field of inquiry before reaching a conclusion.
When evaluating an unfamiliar therapy, we should be willing to ask more than whether one review has declared the evidence conclusive.
- What does the current clinical research show?
- What does the research not yet answer?
- How has the therapy developed historically?
- How is it studied and regulated in different parts of the world?
- Does useful evidence exist in other languages?
- What mechanisms are being investigated?
- What findings support it?
- What findings challenge it?
- Where does genuine uncertainty remain?
A wider investigation does not guarantee a positive answer.
It simply gives us a better chance of arriving at a fair one.
Science Should Leave Room for Curiosity
Science is one of humanity's most powerful tools because it gives us ways to test what we think we know.
It helps us identify bias, measure effects, understand mechanisms and discover risks.
But scientific thinking should also encourage curiosity.
An observation does not suddenly become real on the day somebody publishes a paper about it.
At the same time, an observation does not automatically become true because people have believed it for generations.
Traditional knowledge can provide observations.
Experience can raise questions.
Research can test them.
Science can refine our understanding.
Regulation can establish standards.
Over time, we learn what deserves a place in modern healthcare and what should be left behind.
Perhaps one of the greatest risks in the information age is becoming too certain too early.
As more people begin using artificial intelligence as their first source of health information, this distinction becomes increasingly important.
AI should not merely help us find faster answers.
It should also help us recognise when an answer deserves a wider search.
Because “not yet proven” and “proven not to work” are not the same thing.
Keeping that distinction clear does not weaken science.
It keeps science open to discovery.
Good science should make us more curious, not more narrow.
This article is intended for general educational discussion and does not constitute medical advice. Historical use, regulatory recognition or ongoing research should not be interpreted as proof of efficacy for any specific treatment, product or health claim.
About the Author
Perseus J. Ho is CEO & Managing Director of Bio-Globe Singapore. With more than two decades of experience in the wellness industry, he explores wellness, business, emerging technologies and the evolving relationship between traditional knowledge and modern research.