cbdlovers
EN

CBD side effects: the ones you feel, and the ones you don't

cbdlovers editorial team 13 min

There are two honest answers to this question and they do not overlap at all.

If you are asking what you are likely to notice: tiredness, diarrhoea, and changes in appetite or weight. Every review agrees on that list, including the ones that disagree about everything else. It is a mild list.

If you are asking what safety regulators are actually worried about: the liver, the thyroid and adrenal glands, and reproductive and developmental effects. Not one of those announces itself. You do not feel your liver enzymes rise.

The gap between those two lists is this entire page. Well tolerated and safe are not the same thing, and the difference between them is precisely the set of effects a person cannot detect from the inside.

What you are likely to notice

The uncontested list

Across the clinical literature, the side effects reported often enough to be considered characteristic are consistent and unremarkable:

That is close to the whole of it, at the quantities used in trials — which, it is worth noting, are usually far higher than anything in a consumer product.

Where the list comes from

Two independent reviews, arriving from opposite directions, produce nearly identical lists. The 2017 review in Cannabis and Cannabinoid Research names tiredness, diarrhoea and appetite or weight change as the most commonly reported. The 2019 review in Current Neuropharmacology, which is markedly more cautious in tone, lists diarrhoea, fatigue, vomiting and somnolence in human studies.

When two reviews that disagree about the framing agree about the findings, the findings are probably solid.

Why these particular effects

Diarrhoea in trials is frequently attributable to the carrier oil as much as the compound — MCT oil in quantity has a well-known laxative effect. Somnolence is dose-related and more prominent at the quantities used in epilepsy research than at consumer quantities.

Neither of these is a reason for alarm. They are, however, the entire content of most “side effects” pages, which is why most such pages are reassuring and incomplete.

What regulators are actually worried about

The four systems

EFSA’s 2026 statement on cannabidiol as a novel food identifies concerns about potential adverse effects on the liver, the gastrointestinal tract, the endocrine system, the nervous system and the reproductive system.

That is a substantially longer and more serious list than tiredness and diarrhoea, and the reason it does not appear on most consumer pages is that none of it is something a buyer would ever notice.

Liver

Animal studies showed consistent liver toxicity, with liver weight and histopathological changes identified as sensitive endpoints. Human trial data indicated hepatotoxic potential, particularly when cannabidiol was used alongside other medicines.

This is not theoretical. The prescription cannabidiol medicine authorised in Europe requires liver enzyme monitoring as a condition of use.

Endocrine

Altered thyroid hormone levels and adrenal histopathological changes appear among the animal findings. Thyroid function is not something anyone monitors casually, and its symptoms — fatigue, weight change — overlap uncomfortably with the “mild” list above.

Reproductive and developmental

Animal reproductive toxicity studies reinforced concern on this endpoint, including neurodevelopmental effects following prenatal exposure, with results suggesting long-lasting and sex-specific outcomes. The 2019 review independently lists developmental toxicity, embryo-fetal mortality, reduced spermatogenesis and male reproductive system alterations among animal findings.

Two separate assessments, years apart, landing on the same organ systems, is the kind of agreement worth taking seriously.

The caveat that belongs here

Most of these findings come from animals, and generally at quantities above those used in human therapy. That is a genuine limitation and it should be stated plainly rather than buried.

It is also exactly why EFSA applied an uncertainty factor of 400 — a very large factor, chosen because the data are incomplete rather than because harm was demonstrated at low intake in humans. The resulting figure is a caution level, not a measured threshold.

Why the two lists have nothing in common

Tolerability is not safety

A medicine is well tolerated when people can take it without unpleasant experiences. A medicine is safe when it does not damage them. Those are different properties and a compound can have one without the other.

Everything on the first list is something you feel. Everything on the second is something a blood test finds. A product can therefore be genuinely pleasant to take and still be the subject of open regulatory questions, with no contradiction at all.

Why this favours the marketing

The asymmetry has a commercial consequence. A customer who tries a product and feels fine concludes it is safe, and their conclusion is based on the only evidence available to them. Nothing about that process could have surfaced a liver enzyme trend or a thyroid change.

Which means “millions of people take it and are fine” is a statement about tolerability. It is not evidence about safety, and it never can be.

The interaction problem

This is the one with immediate, real-world consequences, and it is where the risk actually concentrates for most people.

The enzymes

Cannabidiol inhibits several cytochrome P450 enzymes, notably CYP3A4 and CYP2C19. Between them these metabolise a very large share of prescribed medicines.

Inhibiting them means other medicines are cleared more slowly, and their concentrations rise. The effect is the same as taking more of that medicine without deciding to.

Why the epilepsy file proves it

The interaction is not inferred, it is documented. The European assessment file for the prescription cannabidiol medicine describes interactions with clobazam and valproate in detail, and the combination with valproate is specifically associated with liver enzyme elevation.

The 2017 review makes the same point from the other side, noting that more clinical research is needed on cannabidiol’s action on hepatic enzymes and drug transporters — and, notably, that the interaction can cut both ways, potentially reducing the clobazam needed in epilepsy and with it clobazam’s own side effects.

The grapefruit shorthand

If a medication leaflet warns against grapefruit juice, the same enzyme pathway is involved and the same caution applies to cannabidiol. It is crude, it is not a substitute for asking a pharmacist, and it catches most of the important cases.

Who is most exposed

The people most likely to try cannabidiol — for sleep, for anxiety, for pain — are disproportionately people already taking something prescribed. That overlap is the reason this section matters more than the rest of the page.

Who the provisional safe level does not cover

The figure

EFSA’s provisional safe intake for cannabidiol as a novel food, published on 9 February 2026, is 0.0275 mg per kilogram of body weight per day — about 2 mg a day for a 70 kg adult, for highly purified cannabidiol of at least 98 % purity.

The exclusions

The level explicitly does not apply to:

For those groups, EFSA states that the safety picture remains unresolved. The under-25 exclusion is the one that surprises people, and it follows directly from the neurodevelopmental findings.

What this means in a shop

A typical consumer serving is ten to twenty-five times the provisional figure, and the excluded groups together cover a very large share of the people buying. Neither fact appears on any packaging we have seen.

Read the funding line

Here is something the search results hand you directly, if you open the papers rather than the articles about them.

The two reviews Google ranks

For this exact search, Google returns both the 2017 review and the 2019 review among its top ten organic results. They reach different overall verdicts: the 2017 review confirms what it calls the often described favourable safety profile; the 2019 review concludes, in its own words, that cannabidiol is not risk-free.

What the 2017 paper declares about itself

At the foot of the 2017 review is a conflict of interest statement. It says that the European Industrial Hemp Association paid the nova-Institute for the review, and that one author is executive director of the International Association for Cannabinoid Medicines.

The authors disclosed this properly. That is how anyone knows it, and it is to their credit. But the sentence “CBD has a favourable safety profile” has travelled around the world for nine years without that footnote attached, and it is now the most-repeated claim in the industry.

What that does and does not mean

It does not mean the review is wrong. Industry-funded research is not automatically false, the findings on common side effects are corroborated by the independent review, and the paper is transparent about who paid.

It means you should know, when you read a safety verdict, who commissioned it — and that this information is usually available in one click, at the bottom of the abstract.

Compared with what?

The comparator explains the disagreement

The two reviews are not really contradicting each other. They are comparing against different things.

The 2017 review notes that in comparison with other drugs used for these conditions, cannabidiol has a better side effect profile. That is a comparison against antiepileptics and antipsychotics, and against those it is a fair and important claim — one that matters enormously to a family dealing with Dravet syndrome.

The 2019 review compares against nothing at all: is this substance risk-free in absolute terms? No, it is not.

Which comparison you are in

If you are choosing between cannabidiol and an antipsychotic for a serious condition, the first comparison is yours. If you are choosing between cannabidiol and not buying anything, the second one is.

Almost every reader of a page like this is in the second situation, and almost every safety claim they encounter was written for the first.

What is in the bottle is also a safety question

Content that does not match the label

A 2024 review in The Journal of Pain examined products sold in North America and Europe and found cannabidiol content ranging from none to much more than advertised, with other potentially harmful chemicals often present, and serious harm reported in children, adults and the elderly associated with those chemicals.

Synthetic cannabinoids

The most serious documented harms in this market have generally involved products containing synthetic cannabinoids rather than cannabidiol. Those are a different class of compound entirely, with a genuinely dangerous profile, and they end up in products through fraud rather than chemistry.

What a certificate covers

A batch certificate of analysis reports the cannabinoid profile plus panels for heavy metals, pesticides, residual solvents and microbials. Hemp is a known accumulator of heavy metals from soil, which makes that panel more relevant here than for most plants.

Matching the batch code on the certificate to the batch code on the bottle is the only check available to a buyer, and it takes about thirty seconds.

Side effects that are not the cannabidiol

Before attributing anything to the compound, it is worth ruling out the three other things in the bottle. Misattribution is common and it goes in both directions.

The carrier oil

Most oils and softgels deliver cannabidiol dissolved in MCT oil, and MCT in quantity has a well-documented laxative effect. Diarrhoea is on every list of cannabidiol side effects, and a meaningful share of it is the vehicle rather than the passenger.

The test is simple: a product using a different carrier, or a lower volume of the same one, will usually settle it. If the digestive upset follows the oil rather than the cannabidiol, the switch resolves it.

The other actives

A product marketed for sleep that leaves you groggy the next morning is more likely to be reporting the melatonin than the cannabidiol, and CBN, valerian, ashwagandha and 5-HTP each carry their own profiles.

Anything with four ingredients cannot tell you which one produced the effect, good or bad. This is the single strongest argument for buying a plain product when you are trying to learn something.

The hardware, if you vape

Throat irritation and coughing from a vape are usually the device, the temperature or the diluent rather than the cannabinoid. The serious respiratory injuries recorded in 2019 were associated with vitamin E acetate used as a thickening agent in illicit cartridges, not with cannabinoids as such.

That episode is the clearest illustration available of a principle that runs through this whole page: what harmed people was an additive nobody was looking for, in a product bought outside a regulated supply chain.

The trace THC, over time

Full-spectrum products contain THC below the legal ceiling, which is negligible in a single administration and less negligible taken daily into a body that stores it in fat. For anyone subject to workplace screening, repeated full-spectrum use is a different proposition from an occasional one.

What the WHO concluded about dependence

One thing is genuinely settled and it belongs on this page. The World Health Organization’s critical review found that cannabidiol exhibits no effects indicative of abuse or dependence potential in humans.

There is no withdrawal syndrome to plan around and no escalating tolerance documented. Among the things people worry about with a cannabis-derived compound, this is the one that has been answered, and answered reassuringly.

How to judge a source on this topic

The safety query is the one that gets targeted

While measuring the search results for this page, we found that Google’s automated summary for this exact query cites, among its sources, a URL carrying the domain of an academic journal that resolves through a redirect to a commercial page. We found the same pattern on two other cannabidiol searches, on three different institutional domains.

We are not naming them, because those organisations are the victims of the technique rather than its authors. The pattern itself is worth knowing: a trusted domain can be borrowed.

What this means for you

The domain a health claim appears on used to be a decent proxy for its reliability. On this topic it is no longer sufficient. An automated answer assembled from many sources inherits the weakest of them.

Three checks that still work

Look for a named author with an institution. Look for a conflict of interest statement — its absence on a scientific paper is as informative as its content. And look for the primary source: regulator documents and journal abstracts are free, short, and say less than the articles written about them, which is the point.

The short version

What you will probably feel

Tiredness, diarrhoea, appetite and weight change. Mild, consistent across reviews, and mostly at quantities higher than consumer products deliver.

What you will not feel

Liver enzyme changes, thyroid and adrenal effects, and reproductive and developmental findings that have kept a large uncertainty factor in place. These are the reasons the provisional safe intake is 2 mg a day rather than something closer to what is on sale.

What actually matters most

Drug interactions, because the people most drawn to cannabidiol are the most likely to already be taking something that shares its metabolic pathway. A pharmacist resolves this in two minutes.

And who should not rely on any of it

Anyone under 25, pregnant or breastfeeding, or on medication — the three groups EFSA explicitly carved out of its own safe level.