CBD benefits: one is approved, the rest are tiers

Exactly one benefit of cannabidiol has been signed off by a European medicines regulator. It concerns seizures in three named syndromes, it requires a prescription, and it happens at a daily quantity that the food arm of the same institution now calls roughly seven hundred times the provisionally safe intake.

Both of those statements come from official EU files. Neither is wrong. Holding them together at once is the single most useful thing anyone can do with this question, and it is what the rest of this page is for.

Everything else marketed as a benefit sits somewhere below that line, at a level of evidence that can be named precisely. So rather than another list of nine things CBD might do, this page sorts every commonly claimed benefit into tiers, and tells you what would have to happen for one to move up.

Look at what the trusted pages call themselves

Before the evidence, one observation from the search results themselves, because it is unusually honest.

The titles give the game away

The pages that rank for this query are almost entirely hospital and university health systems. Harvard Health calls its piece what we know and what we don’t. UMass Memorial calls its piece fact or fiction. UCLA Health asks considering CBD? and offers what you need to know. Brown Health asks what it is and what it does before it gets to benefits.

Every major institution Google trusts on this topic hedges in the title. The query assumes there are benefits; the trusted answers open by questioning the premise.

And the shortest page ranks

The institutional pages are not long. Several come in under a thousand words, and one of the best-placed is under five hundred. Google is not rewarding depth here. It is rewarding a hospital letterhead on a question where being wrong has consequences.

That is a reasonable instinct on Google’s part and it leaves a gap: nobody is doing the actual sort. Below is the sort.

Tier 1: the one benefit with a regulator’s signature

What Epidyolex is authorised to do

Epidyolex is a cannabidiol medicine authorised by the European Medicines Agency. Its indications are narrow and specific: adjunctive use for seizures associated with Lennox-Gastaut syndrome or Dravet syndrome, in conjunction with clobazam, in patients aged two and over; and seizures associated with tuberous sclerosis complex, alongside other antiepileptic medicines, from age two.

That is the complete list. Three rare syndromes, always alongside another medicine, always by prescription, always under supervision that includes liver monitoring.

Why this is a different kind of fact

Everything else on any benefits list anywhere is at a lower evidentiary tier than this, and the gap is not small. An approval like this one required randomised placebo-controlled trials, a full dossier, a rapporteur, a scientific committee vote, and a European Commission decision. It is the most expensive kind of evidence that exists.

When a jar of gummies implies the same standing, it is borrowing credibility from a file that has nothing to do with it.

What it does not say

The authorisation says nothing about anxiety, sleep, pain, inflammation, skin, or general wellbeing. It does not say cannabidiol is useless for those things. It says nobody has run the file.

The quantity problem, in two official numbers

This is the part that almost never appears on a benefits page, and it reframes the whole question.

What the medicines file describes

The EMA product information for Epidyolex describes starting at 2.5 mg per kilogram of body weight twice a day, rising after a week to 5 mg/kg twice a day, with a ceiling of 10 mg/kg twice daily for Lennox-Gastaut and Dravet, and 12.5 mg/kg twice daily for tuberous sclerosis complex.

For an adult of 70 kg, the upper end of that range works out somewhere between 1,400 and 1,750 mg a day.

What the food file says

On 9 February 2026, EFSA published a provisional safe intake level for cannabidiol as a novel food: 0.0275 mg per kilogram of body weight per day, which for the same 70 kg adult is about 2 mg a day.

Two milligrams. A single supermarket gummy typically contains five to twenty-five times that on its own.

Seven hundred times

Put the two side by side and the arithmetic is brutal:

The one proven benefit occurs at a quantity roughly 700 to 900 times the level European food science currently considers provisionally safe for unsupervised daily use.

The honest caveat on the 2 mg figure

EFSA reached 0.0275 mg/kg by applying an uncertainty factor of 400 to its starting point. That is a large factor, and it exists because the underlying data is incomplete rather than because harm was observed at low intake. The figure is a caution level, not a measured cliff edge, and EFSA calls it provisional in the title of its own statement.

It is also explicitly not applicable to people under 25, to pregnancy or breastfeeding, or to anyone taking medication — for those groups EFSA says the safety picture is simply unresolved.

Why both numbers are right

It would be easy to read the gap as one institution contradicting another. It is not. It is two different questions being asked about the same molecule.

A medicine is judged on benefit against risk

A regulator assessing a medicine weighs harm against benefit in a defined patient with a serious condition. Elevated liver enzymes are an acceptable price when the alternative is uncontrolled seizures in a child with Dravet syndrome, and a neurologist is watching bloodwork throughout.

A food is judged on safety alone

A food regulator has no benefit to put on the other side of the scale. Nobody needs cannabidiol. So the only question is whether it is safe for anyone who might buy it — including the person taking it every day for a decade, with no supervision, alongside whatever else they take.

Under that question, the answer is necessarily far more conservative. That is not timidity; it is the correct output of a different equation.

What that means when you read a benefits list

Any benefit demonstrated at medicine-scale quantities has not been demonstrated at supplement-scale quantities, and the burden is on whoever claims otherwise. This single test disqualifies most of what gets written about cannabidiol, and you can apply it yourself in a sentence: at what quantity was this shown, and how does that compare with what is in the product?

Tier 2: real human trials, no approval

Below the approval line sits a genuinely interesting layer: randomised work in humans that has not been through a regulatory file.

Acute anxiety in experimental settings

The strongest of these are experimental anxiety models — simulated public speaking, for instance — where cannabidiol has produced measurable differences against placebo. These are small, short, often in healthy volunteers rather than diagnosed patients, and they use single administrations at quantities far above any supplement.

They are real evidence. They are also not evidence about taking a 10 mg gummy every evening for six months, which is a different intervention entirely.

Psychosis

There is Phase 2 work examining cannabidiol as an add-on in schizophrenia, with results interesting enough to keep the research programme alive and inconclusive enough that no authorisation followed. This is one of the more scientifically serious threads and it is almost never mentioned in consumer writing, presumably because it does not sell gummies.

Substance use and craving

Several trials have looked at cannabidiol and craving in opioid and tobacco use. Again: real randomised human work, real signals, no approval, and quantities that bear no relation to retail products.

Tier 3: measured while something else was being measured

Sleep

The best systematic review of cannabidiol and insomnia, led from the Mayo Clinic, gathered 34 studies and found that only two focused on patients with insomnia — one of which was a case report. Most measured sleep as a secondary outcome in people recruited for pain or neurological conditions, and the arms that reached significance disproportionately contained THC.

That is what a Tier 3 benefit looks like: not absent, not clean.

Pain

Similar shape. Cannabidiol appears in pain literature mostly as one component of a cannabis-based medicine that also contains THC, or as a secondary outcome. Isolating what the cannabidiol itself contributed is, in most of these papers, not possible.

Why Tier 3 is the most misquoted layer

A study that finds improved sleep scores in a chronic pain trial is a legitimate finding that gets cited as though it were a sleep trial. The citation is technically accurate and the impression it creates is not. When a benefits page links a study, the useful question is what the study set out to measure.

Tier 4: cells and rodents

This is where most of the marketing lives, and it is worth understanding why the tier exists.

Inflammation and neuroprotection

There is a large preclinical literature — cell cultures, mouse and rat models — on cannabidiol and inflammatory pathways, oxidative stress and neuroprotection. It is genuinely interesting to scientists and it is the reason people keep funding trials.

It is also the tier where almost everything fails. Most compounds that look good in a dish never survive contact with a human being.

Skin and acne

Frequently claimed, and traceable to in vitro work on sebocytes. That is a laboratory observation about cells in a plate, not a finding about faces.

The cancer claims

These are the most damaging, and they follow the same pattern: cannabidiol does things to cancer cell lines in laboratory conditions. Extrapolating that to a person is a category error, and one that has led people to delay actual oncology. The WHO critical review is careful and readable on where the evidence sits, and it is the document to read rather than anything written by a seller.

How to spot Tier 4 in the wild

Look for the words in vitro, murine, cell line, animal model, or a study population count that is suspiciously round and large. If the paper never involved a human, the benefit is a hypothesis.

Tier 5: packaging

Claims with no underlying study at all

Immunity. Detox. Balance. Focus. Recovery. Wellness. These appear on labels because they are unfalsifiable, and unfalsifiable is legally safer than specific.

The regulatory tell

There is not a single authorised health claim for cannabidiol in the EU Register of nutrition and health claims. Not one, for any benefit. Under Regulation (EC) No 1924/2006 an unauthorised claim on a food may not be used, whether or not it happens to be true.

So when a product says something vague, the vagueness is doing legal work. A brand that had a defensible specific claim would make it.

How a Tier 4 result ends up on a label

The tiers do not collapse by accident. There is a repeatable chain, and once you have seen it you recognise it everywhere.

The five steps

A laboratory publishes a paper showing that cannabidiol does something to a cell line. The university press office writes a release, and the release drops the words in vitro because they are not reader-friendly. A trade or news site picks up the release and adds a headline in the present tense. A content site cites the news article rather than the paper. A product page cites the content site, or nothing at all, and by then the sentence has become a property of the compound.

Nothing in that chain is a lie. Each step is a small, defensible compression, and five defensible compressions turn a dish of cells into a claim about your body.

Where the information is destroyed

The damage is concentrated at step two, and it is almost always the same two words. In vitro and in mice are the load-bearing qualifications in this literature, and they are exactly what a headline has no room for.

That is why the check that works is not “is there a study” — there is nearly always a study — but “what was the study done in”. A search for the paper title plus the word mice answers it in about fifteen seconds, and it is the fastest way to sort a real finding from a laundered one.

Why sellers are not usually lying

Most people repeating these claims believe them, because they encountered the sentence at step four or five, where it arrived already stripped of its qualifications. Assuming bad faith is both unkind and unnecessary. The chain does the work on its own, and knowing the shape of it protects you without requiring you to think badly of anyone.

Where the entourage effect fits

The argument

Full-spectrum advocates argue that isolated cannabidiol underperforms the whole plant, because minor cannabinoids and terpenes modify its behaviour. It is a plausible pharmacological idea and there is preliminary work behind it.

Why it does not rescue the tiers

The trouble is that the argument is usually deployed to explain away a negative result rather than to predict a positive one. If isolate does not work, the entourage effect is invoked; if it does, nobody mentions it. A claim that survives both outcomes is not doing scientific work.

It also cuts against the marketing in one specific way: if the accompanying compounds matter that much, then the THC in a full-spectrum product is part of the mechanism — and every legal market caps that at a level chosen to keep it from mattering.

What would move a benefit up a tier

For Tier 4 to become Tier 3

A study in actual humans, with the outcome as the thing being measured rather than a side observation.

For Tier 3 to become Tier 2

Randomisation, a placebo arm that is genuinely indistinguishable, validated instruments, and enough participants that the result is not noise.

For Tier 2 to become Tier 1

A full dossier, replication across sites, safety data over long periods, and a regulator willing to put its name on the indication. This is a decade of work and tens of millions of euros, and it is why the tier has one entry.

What you can do with the tiers today

When you meet a benefits claim, ask which tier it comes from and at what quantity it was shown. Two questions, both answerable, and between them they resolve almost every disagreement on this subject.

The short version

What is proven

Cannabidiol reduces seizure frequency in Lennox-Gastaut syndrome, Dravet syndrome and tuberous sclerosis complex, as an add-on, by prescription, at quantities measured in hundreds of milligrams per day and monitored for liver effects.

What is promising

Acute anxiety in experimental conditions, and some work in psychosis and craving. Real trials, no approvals, quantities unlike anything on a shelf.

What is unresolved

Sleep and pain, where cannabidiol has mostly been measured while something else was under study.

What is a hypothesis

Inflammation, neuroprotection, skin and anything oncological, all of which currently rest on cells and animals.

What is packaging

Everything unfalsifiable, and the reason it is unfalsifiable is that a specific claim would be illegal without a dossier nobody has filed.