Tests on St John's Wort tablets showed huge variations in how much active ingredient different brands contain.
St John's Wort is a popular remedy
Some people buying the popular herbal medicine may be getting virtually none, reveals the study by Taiwanese experts.
St John's Wort, or Hypericum, can have mild antidepressant effects, although doctors warn it can interfere with the effects of other medicines.
None of the tablets tested had as much of the active ingredient as claimed, said the researchers.
Two chemicals are thought to be responsible for the psychoactive effects of St John's Wort.
These are hypericin - the chemical ingredient normally named on the label - and a related chemical called pseudohypericin.
None of the five products tested by researchers from the National Chung-Hsing University claimed to contain any pseudohypericin.
In fact, however, this turned out to be the most important active ingredient in the tablets, as they generally contained more pseudohypericin than hypericin itself.
The levels of hypericin fell far short of those mentioned on the label.
On type of tablet contained only 1.7% of the claimed dose - the best brand managed only 38.5%.
Even when the hypericin and pseudohypericin contents were added together, all but one still came up short of the totals on the packaging.
The findings will add to pressure to introduce firmer regulation on the manufacture of herbal medicines - which do not have to be made to the same rigid standards as pharmaceutical products.
Jonathan Berman, from the Health and Safety Group at the Society of Chemistry and industry, said that even if doctors were consulted by patients about possible interactions between their conventional medicines and St John's Wort, it would be impossible for them to know exactly how much hypericin their patient was actually taking.
He said: "Inaccurate labelling has at least one of two effects - the first is potentially to lead to incorrect dose when the label information is complied with.
"The second is to degrade the perceived significance of the label information among either dispensing practioners, or the patients.
"This perception could then carry over to more significant drugs."