"There's castoreum in your vanilla ice cream" — half of that is true, and the wrong half can be checked in the odour field

About 4 minutes.

You have probably seen it: "vanilla-flavoured food actually contains castoreum — the stuff squeezed out of glands next to a beaver's backside."

It usually comes with a photo of a beaver and a line about never being able to unsee it.

The claim splits in two, and the two halves get different answers.

The true half

Castoreum really is a flavour material, and it has a formal number.

Burdock's 2007 safety assessment in the International Journal of Toxicology establishes the identity in its first line:

Castoreum extract (CAS 8023-83-4; FEMA NO. 2261) is a natural product prepared by direct hot-alcohol extraction of castoreum, the dried and macerated castor sac scent glands (and their secretions) from the male or female beaver. It has been used extensively in perfumery and has been added to food as a flavor ingredient for at least 80 years.

The paper also gives its safety conclusions: both FEMA and the FDA regard it as GRAS; acute animal studies show it is nontoxic orally and dermally and neither irritating nor phototoxic to skin; no skin sensitisation was observed in human subject tests; low-level, long-term exposure does not pose a health risk.

So "it is a food additive" and "it comes from a beaver's glands" are both true.

The wrong half

What the legend implies is that it is being used to stand in for vanilla.

That can be checked in the odour field.

The database has eight records with castoreum in the name. I pulled the odor, flavor, odor_descriptions and taste_descriptions fields on all of them and searched for the word vanilla.

Zero hits.

Here is what they actually say:

Record	Suppliers	Odour field
Castoreum absolute	23	leathery, smoky, animal, phenolic, old wood, burnt, resinous
Castoreum resinoid	8	castoreum, animal, leathery, phenolic, resinous
Castoreum liquid (FEMA 2262)	4	castoreum, animal
Castoreum tincture	2	leathery, animal, sweet, smoky
Castoreum extract	0	castoreum, animal

Castoreum absolute's taste description field reads: "leathery animal resinous smoky old wood phenolic burnt".

Nothing sweet, nothing creamy, nothing vanilla.

Their one point of overlap is "phenolic"

Put castoreum absolute's odour field next to vanillin's:

	Odour type	Odour field
Vanillin	vanilla	sweet, vanilla, creamy, chocolate, phenolic
Castoreum absolute	leathery	leathery, smoky, animal, phenolic, old wood, burnt, resinous

Seven words against five, and exactly one in common: phenolic.

That one word makes sense — vanillin is itself a phenol (4-hydroxy-3-methoxybenzaldehyde), and castoreum contains large amounts of phenolic compounds, which is part of where its odour comes from.

So there is a real chemical overlap, and that overlap is "both contain phenols", not "smells like vanilla".

In fragrance description, phenolic means something closer to disinfectant, smoke, medicine and old wood — guaiacol is the representative case, and its flavour field genuinely reads bacon. Nobody reaches for ice cream from that word.

A look at the scale, incidentally

Supplier count is not sales, but it does indicate breadth of coverage:

• Vanillin: 193 suppliers, among the best-supplied materials in this database
• All eight castoreum records together: 60, of which 20 belong to two "replacer" records
• The genuine castoreum records total 40, and the one with a FEMA number — castoreum liquid — has 4

A material with four suppliers cannot underpin the world's vanilla-flavoured food. That fails on order of magnitude alone.

(Burdock's paper does give an annual usage figure, but I can only read the abstract, so I am not repeating that number. I use supplier count as a proxy here, and it is a weak one — why supplier count is not a guide to availability is a separate article.)

So where did the legend come from

I did not research its transmission history, so I can only speak to what I can see:

• "Castoreum is a flavour material" is true. FEMA 2261, 80-plus years.
• "It is used in vanilla flavours" I can find no support for. The database's flavour field files it as leathery type, and its use descriptions run animal, leathery, smoky.
• "Natural flavors" on a label does not itemise constituents, so the claim cannot be falsified — and claims that cannot be falsified travel particularly well.

That is worth remembering: a claim often survives precisely because it lands where the data cannot see.

What to do

• Ask "is it a food additive" and "which food is it in" as separate questions. The first is answerable; the second is not.
• To find a material's use, read the flavor field and flavor_type, not its origin story.
• When you meet a "chemically related" claim, count the shared descriptors. Here it was one out of seven and five.
• Supplier count is fine for an order-of-magnitude check, not for a conclusion.

What this doesn't establish

• I did not investigate what any food company actually uses. This article compares database fields against one safety assessment.
• I could only read Burdock's abstract. Its annual usage figure, exposure estimates and full method are not visible to me, so none of those numbers appear here.
• "Zero records containing vanilla" is the field state of eight records in this database, not the world's fragrance data. Another database may word things differently.
• Descriptor overlap is a crude comparison. Two materials sharing a word need not smell alike, and the reverse holds too.
• Supplier count is not sales, usage or availability. I used it only to check an order of magnitude.
• I did not research the legend's origin or spread.
• "Natural flavors does not itemise" is the general pattern in labelling regulation; I did not check individual jurisdictions' texts.