Almost every herbal guide is organised by plant, which is why almost every herbal guide reads like a catalogue. Organise by mechanism instead and something more useful appears: a handful of ways plant compounds interact with human physiology, each showing up in several body systems at once. Here are seven of them, with each claim graded by what the evidence actually carries.
Ask what pomegranate is for and you will usually be told menopause. Ask what turmeric is for and you will be told joints.
Both answers are too small, and they are too small in the same way. They describe an application and mistake it for the plant's action. Pomegranate does not have a menopause mechanism. It has a receptor-modulation mechanism, and oestrogen receptors are distributed through vascular endothelium, prostate tissue, bone and gut epithelium as well as reproductive tissue. Once you know the mechanism, the list of relevant systems writes itself.
So this piece is organised the other way round. Seven mechanisms, ten plants, and an honest grade on each.
Two of the claims commonly made in this territory do not survive contact with the primary literature. I have kept them in rather than quietly dropping them, because knowing which popular claims fail is worth as much as knowing which hold.
In vitro only · microbiome-dependent
A selective oestrogen receptor modulator is a compound that behaves as an agonist at the receptor in some tissues and an antagonist in others. Tamoxifen is the archetype: antagonist in breast, agonist in bone and endometrium. It is a precise pharmacological term, not a synonym for “plant that does something hormonal.”
Pomegranate itself does not do this. What does — possibly — are the urolithins, which are not present in the fruit at all. They are produced by gut bacteria from the ellagitannins in it.
Larrosa and colleagues, 2006, in Journal of Agricultural and Food Chemistry: urolithins A and B bind both ERα and ERβ, show weak oestrogenic activity in MCF-7 cells, and simultaneously antagonise oestradiol-driven proliferation. Mixed agonist and antagonist activity at the same receptor family — which is, structurally, SERM-like behaviour.
Two caveats the popular version drops. This is cell culture at micromolar concentrations, not humans. And the authors' own framing was cautious — they described urolithins as potential endocrine-disrupting molecules and called for further work. That is a neutral finding, not a therapeutic endorsement.
Then there is the part that makes this genuinely clinically interesting rather than merely plausible.
Not everyone makes urolithins. People fall into urolithin metabotypes — producers of urolithin A, producers of urolithin B and isourolithin A, and a substantial group who produce essentially neither. The determinant is the composition of the gut microbiota, with Gordonibacter among the genera involved, and production happens predominantly in the distal colon.
Which means the same glass of pomegranate juice produces different circulating metabolites in different people, depending on who is living in their large intestine. If the receptor effects turn out to be real in humans, roughly a third of the population would not get them — and that is before we discuss dose.
The cardiovascular and prostate literature on pomegranate is a separate and generally stronger body of work, based on ellagitannins and their antioxidant and endothelial effects, not on receptor modulation. Worth keeping the two arguments apart.
Where I land: calling pomegranate a SERM is a reasonable description of what its bacterial metabolites do in a dish, and an overstatement of anything demonstrated in a person. The metabotype finding is the part actually worth knowing.
Good human evidence · heterogeneous · formulation-dependent
NF-κB is a transcription factor sitting at the centre of the inflammatory response. Activate it and you get coordinated upregulation of TNF-α, IL-1β, IL-6, COX-2 and a range of adhesion molecules. Curcumin inhibits its activation at several points, along with effects on Nrf2, matrix metalloproteinases and the eicosanoid pathway.
This is a different kind of action from an NSAID. An NSAID inhibits a single enzyme family downstream. Curcumin interferes further up, at the point where the inflammatory programme is transcribed.
The human evidence is better than most functional medicine claims and worse than most functional medicine practitioners suggest.
Wang et al. 2021 (Current Rheumatology Reports): sixteen RCTs, 1,810 adults, up to sixteen weeks. Turmeric extracts significantly reduced pain and improved function versus placebo, and produced similar effects to NSAIDs, with about 12% fewer adverse events.
Paultre et al. 2021 (BMJ Open Sport & Exercise Medicine): ten RCTs; all showed improvement from baseline, and the three with direct NSAID comparison found no significant difference in outcome scores.
The caveats that come with those numbers: heterogeneity was very high (I² above 85%), overall risk of bias moderate, every trial short, and BMI significantly modified the effect — benefit fell as BMI rose. No trial has shown structural or imaging change; this is symptomatic relief.
And the point that matters most for an article called food as medicine: these trials used standardised extracts at doses around 500–1,500 mg of curcuminoids daily, frequently with bioavailability enhancement. Culinary turmeric is roughly 3% curcuminoids by weight, and curcumin is poorly absorbed, rapidly conjugated and quickly cleared. Cooking with turmeric is a fine thing to do. It is not the intervention that was tested, and pretending otherwise is the single commonest error in this whole field.
Largely preclinical
Nobiletin and tangeretin, concentrated in citrus peel rather than flesh, are structurally distinct from ordinary flavonoids — the extensive methoxylation makes them considerably more bioavailable and more metabolically stable.
The preclinical work is genuinely interesting: effects on hepatic lipid handling, ApoB secretion, insulin sensitivity and circadian clock proteins, with nobiletin in particular acting on the RORα/β nuclear receptors that gate the clock. In rodent models of metabolic syndrome the results are striking.
Human trial evidence is sparse, small and mostly short. I include these because the mechanism is elegant and worth watching, not because there is a clinical case yet. Eating the zest costs nothing and may deliver something. That is the honest strength of the claim.
Good cardiovascular evidence · weaker antimicrobial evidence
Two distinct things get bundled under garlic, and they have very different evidential standing.
Cardiovascular: this is solid. Ried's 2016 meta-analysis pooled twenty trials and 970 participants, finding a mean systolic reduction of 5.1 mmHg overall, rising to 8.7 mmHg systolic and 6.1 mmHg diastolic in the hypertensive subgroup. Multiple independent meta-analyses agree on the direction, and an umbrella review found seven of eight meta-analyses showing a substantial systolic effect.
The honest caveat comes from Rohner and colleagues in the American Journal of Hypertension: heterogeneity is high (I² = 71%), and when analysis was restricted to higher-quality trials with concealed allocation and blinded measurement, the effect size fell. Real, then, but probably smaller than the headline figures — and a first-line antihypertensive remains a first-line antihypertensive.
Antimicrobial: allicin, formed when alliin meets alliinase on crushing, has broad in vitro activity against bacteria, fungi and some viruses, working through thiol-group interference — a non-specific mechanism to which resistance is harder to develop. It is also unstable, degrading within hours, and poorly represented in most supplements. In vitro potency has not translated into demonstrated systemic antimicrobial efficacy in humans. Reasonable adjunct; not an antibiotic.
Nobel-winning drug · the plant is a different question
Artemisinin, isolated from Artemisia annua by Tu Youyou, earned the 2015 Nobel Prize in Physiology or Medicine and has saved an enormous number of lives. The endoperoxide bridge generates reactive species on contact with iron, which is why parasites concentrating haem are selectively vulnerable. Artemisinin-based combination therapy is the global standard for falciparum malaria.
None of which licenses the use it gets in functional medicine.
The drug is not the plant. The WHO explicitly advises against non-pharmaceutical Artemisia annua preparations — teas, dried leaf, herbal capsules — for malaria treatment or prevention. Variable and sub-therapeutic dosing is a recognised route to artemisinin resistance, and resistance to this compound class is a genuine global health concern.
Antimalarial is not antiparasitic-in-general. Evidence for artemisinin or wormwood against the intestinal parasites and dysbiotic patterns it gets prescribed for in practice is thin. Artemisia absinthium has a separate traditional anthelmintic history and a separate, smaller evidence base — and thujone toxicity constrains dosing.
Where a GI-MAP shows an actual identified parasite, the appropriate response is targeted treatment, frequently pharmaceutical, with retesting. Not a botanical protocol chosen because the mechanism sounds right.
In vitro · weakest claim in this article
Biofilm is a real and important problem. Bacteria within an extracellular polymeric matrix can be orders of magnitude less susceptible to antimicrobials than the same organisms free-floating, and biofilm is genuinely implicated in persistent infection.
Cistus incanus is polyphenol-rich and shows anti-adhesive activity in vitro — interfering with bacterial surface attachment, which is the first step of biofilm formation. There is some human work on oropharyngeal bacterial load and on Borrelia in laboratory culture.
I am including this to be complete and to be clear about where it sits. In vitro anti-adhesion in a dish is a long way from disrupting established biofilm in a human gut, and the leap between those two claims is made routinely and without evidence. If a practitioner tells you a herbal preparation is breaking down your biofilm, ask what measurement supports that. There generally is not one.
Cistus tea is pleasant and safe. That is the level of claim the evidence supports.
Rodent data only · no human trials in either direction
This one is worth walking through carefully, because it is repeated with a confidence that the source material does not support anywhere.
What exists is rodent work. Aga and colleagues, 2001, in Journal of Ethnopharmacology: mice given lead acetate in drinking water, with coriander administered by gastric intubation, showed reduced lead deposition in the femur and less renal injury. Téllez-López and colleagues, 2017, found methanol extract lowered lead concentration and protected liver tissue in Wistar rats. Mustafa, 2021, found similar protection in rat neural tissue.
Three points about that body of work.
It is entirely animal. There are no human trials of coriander for heavy metal burden. None.
These are prevention models, not mobilisation models. In each case coriander was given alongside or shortly after ongoing exposure. That is a different question from whether it mobilises metal already stored in bone or brain, which is the claim usually made clinically.
The Téllez-López paper opens by saying so. Its stated rationale is that coriander is said in traditional medicine to have chelating properties but that scientific evidence to support this is absent. That is the authors' own framing of the field they were entering.
The standard functional medicine warning — that cilantro mobilises metals without binding them and can therefore redistribute them, particularly to the brain, unless a binder is given alongside — is itself an unevidenced claim. It traces to clinical anecdote rather than to any trial.
So the position is symmetrical and slightly unsatisfying: no human evidence that cilantro mobilises stored metal, and no human evidence that it causes harmful redistribution. Both confident statements are running well ahead of the data. Where genuine heavy metal toxicity is established, chelation is a medical intervention with monitoring, not a food.
Eat coriander because it is delicious and a decent source of polyphenols. That is what the evidence supports.
Small short RCTs · consistent direction
Melissa officinalis works through a mechanism specific enough to be worth naming: rosmarinic acid inhibits GABA transaminase, the enzyme that degrades GABA. Inhibit the breakdown enzyme and synaptic GABA persists longer. That is a defined pharmacological action rather than a vague calming property, and it is broadly the same strategy as several anticonvulsants, at a far lower intensity.
Human trials exist and are consistently in the right direction — reduced anxiety ratings, improved sleep quality, better calmness scores under induced stress, including some work in people with cardiac conditions and palpitations. They are also small, short, and frequently conducted by groups with a product interest.
On the HPA axis specifically: the honest position is that the effect is on subjective anxiety and autonomic arousal, with cortisol outcomes measured inconsistently and reported inconsistently. Claiming lemon balm regulates the HPA axis goes further than the trials do.
What I would say clinically: it is safe, it is cheap, it is pleasant as a tea, the mechanism is real, and it does something modest for some people. That is a perfectly respectable claim and does not need inflating.
Small effect · every trial manufacturer-funded
Rose hip is usually presented as a vitamin C story: exceptional vitamin C content, therefore collagen synthesis, therefore connective tissue. The first part is true of the fresh fruit. The rest needs adjusting.
Vitamin C is genuinely required for connective tissue — it is the cofactor for prolyl and lysyl hydroxylase, the enzymes that hydroxylate proline and lysine residues in procollagen. Without it, collagen triple helices are unstable. That is why scurvy presents as connective tissue failure. This part is settled biochemistry.
But the vitamin C is not what the rose hip trials were testing. The compound credited in that literature is a galactolipid, GOPO, with anti-inflammatory activity independent of ascorbate. And vitamin C is heat- and oxidation-labile, so content varies enormously with cultivar, harvest timing, drying and processing. A dried commercial rose hip powder may contain a fraction of the fresh figure.
The clinical evidence: Christensen and colleagues, 2008, in Osteoarthritis and Cartilage, pooled three RCTs and 287 patients. Rose hip powder reduced pain with an effect size of 0.37 (95% CI 0.13–0.60), number needed to treat of six, and reassuringly low heterogeneity.
And the caveat the authors state plainly: all three trials were supported by the manufacturer. They describe the evidence base as sparse and call explicitly for independent replication in a larger, longer trial. Eighteen years on, that replication has not really arrived.
A small, consistent effect from three industry-funded studies is not nothing. It is also not a foundation to build a connective tissue protocol on, and the mechanism people cite for it is probably the wrong one.
Does not hold up — see below
This claim appears in almost every longevity-adjacent herbal list, and I have repeated it myself in the past. Having gone back to the primary literature, I do not think it stands.
Four problems, and they compound.
The compound is not the herb. The telomerase-activating molecule is cycloastragenol, the aglycone of astragaloside IV — a purified, concentrated derivative sold as TA-65. Astragalus root, decoction or standard extract is not that, and no evidence suggests it delivers a meaningful dose of it.
The mechanistic work is cell culture. Telomerase activation was demonstrated in keratinocytes, PC12 cells and primary rat neurons. Not in people.
The pharmacokinetics are poor. Zhu and colleagues, 2010, found that after thirty minutes in human liver microsomes, only 8.2% of starting cycloastragenol remained. The authors concluded that extensive first-pass hepatic metabolism would limit oral bioavailability. That is the paper's own conclusion, in a study by the group developing the compound.
The single human RCT has a fatal internal inconsistency. Salvador et al., 2016, in Rejuvenation Research: 117 CMV-positive adults aged 53–87, twelve months. The low dose (250 U) increased telomere length by 530 bp while placebo lost 290 bp. The high dose (1,000 U) did not reach significance.
That last point deserves emphasis. If the mechanism were genuine telomerase activation, four times the dose should not do less. A non-monotonic dose-response with significance only at the low dose is the classic signature of a chance finding rather than a drug effect.
Add the conflicts: authors affiliated with T.A. Sciences, the manufacturer; Calvin Harley, formerly of Geron and central to the compound's commercial development; publication in a journal founded specifically to advance the longevity field. That is not automatically disqualifying, but with a single trial and an incoherent dose-response, independent replication would be the minimum.
There is also a plausible alternative explanation on the table. In a CMV-positive cohort, leukocyte telomere length depends heavily on the mix of cell types sampled. A shift toward naive T cells, which have longer telomeres, would raise measured average telomere length without any telomere having lengthened. Immune redistribution, not rejuvenation.
What I would say instead: astragalus has a reasonable case as an immune-supportive adaptogen with a long traditional record and some supportive immunological data. That is a defensible claim. Telomere lengthening is not, and continuing to say it costs credibility that the defensible claims then have to pay for.
The argument usually offered to unify all of this is that what the body cannot synthesise it obtains from food, and that plants therefore contain what we need by design.
Half of that is right, and it is the important half. We genuinely cannot make vitamin C, or the essential amino acids, or the essential fatty acids, and we obtain them from food or we fail. That is a real and load-bearing argument for eating actual plants rather than isolated compounds.
The design part I would drop, and I say that as someone who finds it rhetorically appealing.
Plants do not make these compounds for us. Curcumin, allicin, artemisinin and the polyphenols are largely defence chemistry — antimicrobial, antifungal, insect-deterrent, produced under stress. That they do interesting things in human tissue is a happy consequence of shared biochemistry and a long history of co-evolution with what we ate, not evidence of intent. The same defence chemistry gives us oxalates, lectins, cyanogenic glycosides and thujone.
Which is actually a more useful frame, because it predicts things the design argument does not: that dose matters enormously, that the same compound can be therapeutic and toxic, and that traditional preparation methods — soaking, fermenting, boiling, discarding the water — are usually solving a real problem rather than performing a ritual.
Ranked by what I would actually rely on:
Notice that the right-hand column is not a list of useless plants. It is a list of claims that outran their evidence — and in most cases the same plant has a smaller, better-supported claim sitting right next to the inflated one.
That is the recurring shape of this field. Not fabrication. Escalation. A mechanism found in cell culture becomes a property of the compound, then a property of the plant, then a clinical indication, and by the time it reaches a supplement label nobody has checked which step it stopped being true at.
Checking is not scepticism about plants. I use all ten of these. It is just insisting that what gets said about them is the same size as what has been shown.
A one-page companion table — all ten plants, their active compound, mechanism, evidence grade, and the specific caveat attached to each — is available alongside this article for printing or clinic use.
Educational content, not medical advice. Several of these plants interact with medication — garlic with anticoagulants and some antiretrovirals, turmeric with anticoagulants and at high doses with iron absorption, and the artemisinins are prescription antimalarials with their own resistance considerations. Check before combining anything here with prescribed treatment.
Which of these is relevant to you depends on what is actually happening — including, in the case of pomegranate, which bacteria you are carrying.
Ask AIdan → See the TDG Programme