Cordyceps human research: what has actually been studied?

If you know one thing about Cordyceps, it is probably that it grows out of an insect. That description sounds like somebody trying to make a mushroom more interesting than it really is, but in this case it is basically what happens. The fungus infects an insect, grows through it, kills it and eventually produces a fruiting body from the remains. In the famous Tibetan caterpillar fungus the whole thing is collected together, dead caterpillar underneath and fungal stalk growing out of it.

I have always thought fungi were strange, but Cordyceps does make quite a strong opening argument.

The complication is that we then took the word Cordyceps and started using it for several rather different things. The wild caterpillar fungus with the enormous traditional reputation is now called Ophiocordyceps sinensis. The bright orange cultivated fungus in most of the photographs is usually Cordyceps militaris. Then there are mycelial fermentation products such as CS-4, Jinshuibao, Cordymax, Corbrin and Bailing, some of which have moved through several fungal names over the years. Cordycepin gets bundled into the same story as well, despite being a single compound first isolated from C. militaris.

So before we get anywhere near whether Cordyceps “works”, we have the same problem that keeps appearing in mushroom research.

What exactly are we talking about?

Winter worm, summer grass

The organism at the centre of the traditional Cordyceps story is Ophiocordyceps sinensis, formerly called Cordyceps sinensis. It lives in high-altitude regions of the Tibetan Plateau and Himalayas and parasitises ghost moth larvae living underground.

The larva becomes infected, the fungus grows through the body and eventually produces a dark fruiting structure from around the head. What gets collected and sold is the two organisms joined together, which is why photographs of wild Cordyceps look so different from the orange cultivated mushrooms most people now see online.

The Tibetan name yartsa gunbu and the Chinese dong chong xia cao are both usually translated along the lines of winter worm, summer grass. You can see exactly how that name arose. For part of the year there appears to be a worm in the ground, and then something plant-like grows out of it. Nobody needed to know about fungal taxonomy to notice that something very peculiar was going on.

The host story is not especially tidy either. A review examined 91 insect names across 13 genera associated with O. sinensis and judged 57 to be potential hosts, eight indeterminate and 26 not hosts. Even the idea of one neat “Cordyceps caterpillar” falls apart once somebody starts checking properly.

This is one of the things I like about fungi. We give them a simple description because we need one, and then the organism gets on with being considerably more complicated.

Wild Cordyceps

Wild Ophiocordyceps sinensis is rare, expensive and economically important in the regions where it is collected. Fieldwork in Dolpa, Nepal found the trade contributing more than half of household cash income at one point, with poorer households particularly dependent on it. At the same time prices rose dramatically, individual harvests fell and most collectors reported that availability was declining.

That gives you some idea why wild Cordyceps is not what fills the shelves of the supplement industry.

Most modern products use cultivated or fermented material. Some use Cordyceps militaris, which can be farmed much more readily. Others use fungal biomass grown by submerged fermentation and sold under names linked historically with O. sinensis. Once you realise this, a lot of the modern market makes more sense.

It also raises a question that comes up constantly in mushroom research: if the species or preparation changes, how different is the finished material?

With Cordyceps we know the answer is at least “different enough to measure”. One metabolomics study found 368 metabolites that differed between wild and cultivated O. sinensis. Another study using Raman spectroscopy and machine learning could distinguish wild from cultivated material with about 99% accuracy.

That still does not tell us which is better. In fact the nucleoside work manages to argue with itself. One study found cultured mycelium richer in several nucleosides than natural material, while later work found many nucleosides and nucleotides higher in wild material. Different strains, different samples, different methods, different answers.

Fine. That is science behaving like science rather than a sales sheet.

What we can say is that wild, cultivated and fermented material should not be treated as though nothing changes between them.

And then there is Cordyceps militaris

Cordyceps militaris is the bright orange species that appears in grow kits, photographs and a lot of modern Western products. It is a different species from Ophiocordyceps sinensis, much easier to cultivate, and interesting in its own right.

It is also where cordycepin enters the story.

Cordycepin was isolated from C. militaris cultures in 1950 and later identified as 3′-deoxyadenosine. It looks chemically similar to adenosine but is missing one hydroxyl group, which turns out to make it biologically interesting enough that researchers have been studying it ever since.

The trouble is that the whole Cordyceps story then gets folded together. A page starts with a Tibetan caterpillar fungus, moves into cordycepin, and ends by selling C. militaris. You are left with the impression that somebody has simply followed one organism from traditional medicine into a modern laboratory.

They have not.

The wild caterpillar fungus, C. militaris and the various fermentation products are related parts of the Cordyceps story, but they are not interchangeable.

The fermentation problem

A large part of the human Cordyceps literature is not about a mushroom fruiting body at all. It is about fermented fungal material.

CS-4 is probably the best known example. Products such as Jinshuibao, Cordymax and Corbrin sit around the same area of the literature. Rather than harvesting wild caterpillar fungus, fungal cultures are grown under controlled conditions and the resulting biomass or extract is used as the product.

Then there is Bailing, which manages to make the naming problem particularly obvious. One trial calls the organism in Bailing Hirsutella sinensis. Other papers describe material associated with the same product as Paecilomyces hepiali. Fungal nomenclature changed again when the one-fungus-one-name rule came into effect, retiring separate anamorph names such as Hirsutella sinensis.

So you can read one paper saying Cordyceps sinensis, another saying Hirsutella sinensis, another saying Paecilomyces hepiali, and then look at a commercial product simply called Cordyceps.

This is not a minor technical problem hidden in the methods section. Ten of the 34 human studies in our current library do not give enough information for us to place the tested material confidently into a modern material group.

I think that is one of the most important things to understand before reading the clinical evidence. The word on the front of the packet is often doing far more work than it should.

Where did people start using it?

The standard supplement story usually says Cordyceps has been used in Traditional Chinese Medicine for centuries. There is certainly Chinese use and Chinese medical literature around it, but the history we have found so far points just as strongly towards Tibetan and wider Himalayan traditions.

Yartsa gunbu is embedded in Himalayan medicine, trade and culture. Fieldwork in Dolpa recorded use by collectors and amchis for a wide range of complaints, as well as more general tonic and aphrodisiac use. In Manang, traditional practitioners were recorded using the fungus in medicines themselves rather than simply collecting it for sale. North Sikkim material describes an extremely broad medicinal reputation, and a Naxi market survey in Lijiang recorded chong cao as a widely recognised medicinal material.

These records are useful because they show us what people actually did with the organism and how they understood it. They do not need to be turned into clinical evidence to be interesting.

There is a widely repeated claim that the earliest written account is in a fifteenth-century Tibetan medical text associated with Zurkhar Nyamnyi Dorje. At the moment we have not verified that from the primary text or the specialist scholarship needed to settle it properly. So it stays as an interesting claim rather than something I am going to confidently date on a public page.

This is exactly the sort of historical detail that gets polished with every retelling. A modern author writes “possibly”, somebody else removes the word, and ten years later a supplement company has an exact date and a monk.

We can do better than that.

Ayurveda

Cordyceps is also a useful example of how easily Indian and Himalayan traditions get flattened into Ayurveda.

We searched a large tradition corpus and did not find a classical Ayurvedic source for the caterpillar fungus. What we did find was Himalayan ethnomedicine, Tibetan medicine and local traditions from parts of India and Nepal. Some papers even appear in journals with Ayurveda in the title, which is probably how the story starts to drift.

A journal title is not provenance.

If a genuine classical Ayurvedic source turns up, I will happily put it in the library. At the moment we do not have one, so I am not going to pretend we do.

Homeopathy

This is almost the opposite problem.

When I wrote my fungi dissertation in 2022, Cordyceps stood out because it appeared much more clearly in the homeopathic fungal literature than Lion’s Mane or Reishi. The latter two were barely represented compared with the older fungal remedies such as Agaricus, Bovista and Claviceps, whereas Cordyceps actually turned up as something worth discussing.

We have not yet completed the proper dedicated homeopathic search for this research library, so I do not want to lift a few old notes into this page and pretend the job is done. That work is next.

I am interested to see what survives when we go back to the materia medica properly, because it is another example of different traditions noticing completely different fungi.

What is actually in Cordyceps?

This gets awkward because, once again, the answer depends on which Cordyceps you mean.

The wild caterpillar fungus, cultivated O. sinensis, fermented mycelium and C. militaris do not have identical chemical profiles. Polysaccharides, nucleosides, sterols, proteins and other compounds all appear across the literature, but their amounts and structures vary with species, strain, substrate, growth conditions and processing.

Cordycepin is the compound everybody knows because the name is memorable and the laboratory literature is enormous. It is 3′-deoxyadenosine and was isolated from C. militaris, not from some mythical universal Cordyceps essence.

There is also a pharmacology problem that makes the story rather less convenient. Human adenosine deaminase breaks cordycepin down extensively, which is why pharmaceutical researchers often study it with an inhibitor of that enzyme. A compound doing something dramatic in a dish is therefore a starting point, not evidence that eating a mushroom produces the same exposure in a person.

In the human research we have collected, nobody was given purified cordycepin as the intervention.

That matters because cordycepin often gets used as the scientific explanation for whatever somebody wants to say about a Cordyceps supplement. At the moment the human evidence does not support making that jump.

Laboratory and animal work

There is a lot of preclinical Cordyceps research, and some of it is genuinely fascinating.

Researchers have compared wild and cultivated materials, mapped metabolites, sequenced genomes, looked at polysaccharides, studied cordycepin metabolism, investigated antioxidant and inflammatory pathways and tested C. militaris in animal models. There is enough material to keep a lab busy for a very long time.

What I found particularly useful was that even the preclinical literature keeps pointing back to the material problem. A paper comparing cultured C. militaris with natural C. sinensis prompted a published comment and rebuttal about whether the comparison was sensible in the first place. Two nucleoside studies can point in different directions depending on which samples were tested. Reviews of Cordyceps polysaccharides admit that the relationship between structure and biological activity is still not well established.

This is not a failure. It is what happens when a complicated biological material gets reduced to one commercial word.

The laboratory work gives us mechanisms, questions and possible leads. It does not mean that a person swallowing a Cordyceps capsule experiences whatever happened to a cell or rat.

We keep all of it because throwing away the preclinical story would be ridiculous. We just keep it on the correct shelf.

The human research

There are currently 34 verified human primary studies in our Cordyceps library. Across the studies where a participant number can be established under our counting rules, that gives 2,392 people. Fifteen studies were placebo-controlled, covering 838 participants.

That sounds like quite a lot until you look at the material.

There are no human studies of wild Ophiocordyceps sinensis in the corpus. There are no human studies of cultivated O. sinensis fruiting body either.

I had expected that number to be small. I did not expect it to be zero.

Every human study we found used a fermentation product, Cordyceps militaris, another Cordyceps taxon, or material that the paper did not identify well enough for us to place properly. Nine studies used fermented mycelial material referred to O. sinensis. Six used Bailing-type Paecilomyces/Samsoniella hepiali fermentation. Six used C. militaris. Three used other Cordyceps taxa and ten could not be placed confidently.

That changes how you read the whole subject.

The organism carrying the traditional reputation, the one people picture when they talk about Tibetan Cordyceps, is not the thing that has generated the modern clinical literature.

Of the 34 studies, 26 were intended to test efficacy. Under the rules we use throughout this library, 11 are positive, six mixed, one null, one negative, seven unresolved from the available source and eight are not really efficacy studies. Only 19 of the 34 give us a readable efficacy direction at all, and only seven of those 19 were double-blind.

There is human research here. It is simply nowhere near as clean as the word Cordyceps makes it sound.

The human research is not mainly about energy

This surprised me a little because the modern Western Cordyceps story is so dominated by energy, endurance and exercise.

The clinical literature is much more occupied with kidneys, kidney transplantation, chronic respiratory disease, immune measurements and various Chinese medical preparations. A lot of that work uses fermentation products rather than fruiting bodies.

One of the better single trials in the corpus was published in 2024. It randomised 240 people with chronic respiratory disease, analysed 238 and followed them for 48 weeks. The Bailing preparation beat placebo on exacerbations and adverse-event rates were very similar between groups.

That is a result worth paying attention to.

It is also a result for a fermentation product described in the paper as Hirsutella sinensis. It is not a trial of wild caterpillar fungus and it is not a trial of Cordyceps militaris.

A 2023 multicentre study of Jinshuibao in 200 people recovering from COVID-19 also reported better improvement in several patient-reported cardiopulmonary symptoms after two weeks. Again, interesting. The abstract does not identify the organism beyond describing the product as a Traditional Chinese Medicine, and it does not even give us a dose.

This sort of thing happens throughout the Cordyceps literature. You find an interesting result and then spend the next hour trying to work out what fungus was actually in the capsule.

What about energy and endurance?

This is probably the section most people expected to be much more impressive.

The cleanest placebo-controlled exercise study we found gave 22 trained male cyclists 3 g a day of Cordymax CS-4 for five weeks. It found no difference in VO2peak, ventilatory threshold or time-trial performance.

Another study in twenty healthy older adults is cited constantly because the treated group showed a 10.5% increase in metabolic threshold. It sounds much better until you notice that the 10.5% is a before-and-after change within the Cordyceps group. The paper did not establish that change against its own placebo group.

That is exactly the sort of thing our research rules are designed to catch. A group improving after an intervention is interesting. The intervention beating the control is what tells you whether the improvement is likely to be attributable to it.

There are a few other small studies involving runners, young adults and acute exercise. Some are mixed, some unresolved and some measure what exercise did rather than what the supplement did.

So the sports story is not empty, but it certainly does not justify the certainty with which Cordyceps is often sold for energy and endurance.

People may still take it and feel more energetic. I have no difficulty believing that people report that. It is simply not the same question as whether the controlled trials have demonstrated an effect.

What do the reviews make of it?

We currently have 15 systematic reviews and meta-analyses in the Cordyceps library. Quite a few concern kidney disease or dialysis, with others looking at COPD, kidney transplantation, lung cancer, Hashimoto’s thyroiditis and broader bodies of Cordyceps research.

As with Reishi and Lion’s Mane, the review label can make the literature sound rather grander than it is.

A 2023 diabetic kidney disease review included 38 trials but reported that none provided outcomes on kidney-disease progression or cardiovascular events. A 2019 COPD review included 15 trials and 1,238 people but said none was placebo-controlled or of high methodological quality. A 2024 umbrella review looked across 14 systematic reviews and rated 13 of them critically low quality.

There may be signals worth following, especially around some fermented products, but a pile of weak trials does not become a strong trial merely because somebody runs a meta-analysis over the top.

The material problem appears here too. Ten of the 15 reviews cover material that cannot be resolved to one clear organism or preparation, and several combine different Cordyceps materials.

You can see why I keep going on about the name.

Safety

Cordyceps safety turns out to contain several completely different stories that would be easy to muddle together if we just wrote “Cordyceps adverse events”.

The first is contamination of wild O. sinensis. One report found lead at an extraordinary concentration in a contaminated product associated with two poisoned patients. Arsenic has also been found above limits in many tested wild batches and appears to be concentrated particularly in the insect part of the caterpillar-fungus complex.

Even there the answer is not simple. Much of that arsenic has not been chemically identified, and different risk assessments come to different conclusions depending on what they assume about the unknown fraction.

The second story involves wild cicada fungi eaten as food in Vietnam and Thailand. Across the reports in our corpus there were 99 poisoning cases and one death associated with Ophiocordyceps heteropoda and ibotenic acid. That is serious, but it is not a safety signal for the cultivated Cordyceps supplement somebody has bought from a shop in Wrexham. Different fungus, different material, different exposure.

The third story is more relevant to modern commercial products. A Thai case series reported 11 cases of acute kidney injury associated with Cordyceps militaris, including two people who required dialysis. At the moment that report is still a preprint, so it stays labelled as one. There are also individual reports involving bleeding, hepatoportal sclerosis and occupational allergy to another Cordyceps-related fungus.

These records matter, but case reports do not tell us incidence. They tell us what has happened, not how often it happens.

So, does Cordyceps work?

This is another question that becomes less useful the more you learn about the mushroom.

There are positive human trials. Some fermented preparations have produced interesting results in specific populations, particularly in parts of the respiratory and renal literature. There are also mixed studies, a null exercise trial, a negative insomnia result, unresolved papers and an enormous amount of research whose material identity is less clear than I would like.

What we do not have is human trial evidence for the famous wild caterpillar fungus itself.

That is quite a remarkable thing when you think about it. The organism carrying most of the traditional reputation has no human study in our current corpus, while the clinical literature everybody points to is largely about cultured fungal products developed much later.

The modern energy claim is not especially convincing either. The neatest athlete study was null and one of the most quoted positive exercise results is a within-group change that was never shown against placebo.

So I would not say “Cordyceps works” and I would not say “Cordyceps does nothing”. Both sentences flatten several species, preparations and bodies of evidence into something much simpler than they really are.

I also do not think the clinical uncertainty makes the subject disappointing.

You have a fungus that parasitises insects, a centuries-old Himalayan trade, a traditional medical history that still needs proper primary-source work, a wild resource worth extraordinary amounts of money, cultivated orange fruiting bodies, fermentation products with their own clinical literature, and a compound such as cordycepin that chemists have been pulling apart for more than seventy years.

That is already interesting enough.

The clinical evidence should be allowed to be exactly what it is, rather than being asked to provide the reason the mushroom matters.

What is still unknown?

Quite a lot, which is becoming a familiar ending to these pages.

We still need better access to the specialist Tibetan historical literature before making confident statements about the earliest written sources. We have not found a verified classical Ayurvedic source. The homeopathic literature needs its own dedicated search rather than being represented by what I happened to collect for my dissertation.

Three human studies are still held at metadata level because we have not obtained an abstract or full text, and 18 relevant registered trials currently have no publication and no posted results.

The material problem is probably the biggest scientific gap. We need better trials that say exactly what organism, strain, fungal part and preparation were used. We also need human studies of the materials consumers actually think they are buying when they see the word Cordyceps.

And, rather wonderfully, after all this research there is still one very obvious study missing.

Nobody in our corpus has actually run a human trial on the famous wild caterpillar fungus.

The complete Cordyceps research library

Research last searched: 3 October 2026

The current library contains 34 verified human primary studies involving 2,392 participants under our counting rules, including 15 placebo-controlled studies involving 838 people. All 34 human studies have had two independent verification passes and none currently needs review. The library also contains 15 systematic reviews or meta-analyses, 16 selected preclinical records, 13 safety records and 11 traditional or historical records.

The full tables below keep the different Cordyceps materials separate rather than pretending a fermented mycelial product, an orange C. militaris fruiting body and a wild O. sinensis caterpillar fungus are one interchangeable thing. They also keep positive, mixed, null, negative and unresolved findings together.

References