The placebo is one of science's greatest mysteries. The pill that isn't a pill. The medical illusion that somehow becomes real.

The mind-boggling weirdness of the placebo effect is certainly a strange thing, nobody doubts that.

But just because the placebo effect occasionally delivers unexpected outcomes doesn't mean we should overestimate how powerful it is – nor try to find a place for it in the medical care of patients, scientists are now warning.

In a new perspective article, researchers from the University of Sydney argue that recent suggestions placebos could play a role in clinical care are unfounded, and are based on flawed evidence.

"Much of the current discourse on placebo seems to focus more on enshrining placebos as mysterious and highly effective and less on making a practical difference to patient care and outcomes," a team led by first author and physiotherapy researcher Chris Maher writes in the commentary.

Observations of the placebo effect can be traced back to the 18th century, and the reputation of the placebo has grown ever since: the idea that an inert, sham treatment, taken unknowingly by a patient, can sometimes deliver therapeutic effects like the real thing.

That reputation is mostly underserved, Maher and his co-authors say, noting that if you look closely at much of the evidence, placebos tend to only offer very modest effects, and even then only in a small minority.

"A Cochrane review of placebos considered 234 trials and concluded that, in general, placebos do not produce major health benefits, except for some small and inconsistent effects on self-reported outcomes such as pain or nausea," the researchers explain.

Not everybody holds this view, however.

In recent times, a number of commentators have suggested placebos might be worth exploring as substitutes for use in the medical care of patients, on the basis that we might be able to exploit the placebo effect for clinical outcomes.

In their new article, Maher and colleagues list some of the problems with these arguments, pointing out that a number of the experiments investigating the placebo effect have significant limitations to bear in mind – such as very small samples of participants – or have flawed designs that don't make for strong evidence or practical application.

For example, some studies repeatedly primed patients on the effectiveness of placebos for lengthy periods, which would not be feasible during a typical short medical appointment.

None of this is to say that the placebo effect itself isn't an observable phenomenon.

But until there's real, firm data to suggest otherwise, placebos have no place being given to patients in clinical scenarios, the researchers argue – not when actual medicine is an option on the table.

"Placebos remain important for clinical trials because they help achieve blinding and, thus, control of bias," the team writes.

"When administered in a blinded fashion, a placebo will provide a small effect, but the real treatment will normally provide better outcomes for the patient… It may be better to dismiss placebos and instead manage patients with evidence-based treatments."

The perspective is published in the Medical Journal of Australia.