We found one plant sold under 50 different names

Tahoe, Larry, SFV, Skywalker#

Walk into a dispensary and ask for Tahoe OG, Larry OG, SFV OG or Skywalker. You'd expect four different strains, four different stories and four very confident budtender monologues. In the DNA, they measure as one plant wearing four name tags.

So do Ghost OG, Hitman OG, Khalifa Kush, Kosher Kush, 707 Headband, Triangle Kush and forty more. In total, 63 samples sold under 50 names line up as a single genome. Picture a group photo of 63 people where everyone has the same face. And we didn't cheat with a daisy chain of lookalikes. We compared every sample with every other sample in the group, and every pair passed.

That's the kind of thing you only see when you put thousands of DNA results next to the names they were sold under. CannaLib holds genotypes (a plant's DNA fingerprint) for 3,584 samples from four public datasets, matched to its catalogue of 143,611 strains. This piece uses the largest panel: 2,264 samples genotyped by Phylos Bioscience and published in the European Nucleotide Archive [1].

What "the same plant" means#

Most cannabis in shops is grown from cuttings, not seeds. A cutting is a photocopy: snip a branch off a mother plant, root it, and the new plant has exactly the same genome. Pass that cut from grower to grower and it stays the same plant, even if every grower gives it a new name. Two samples of one cut measure as near-identical twins.

To spot them, we use the standard kinship test geneticists use to find duplicates in DNA studies [2]. Think of it as a family-resemblance meter that runs from 0 (as alike as two strangers) up to about 0.5 (the same plant, twice). A score above 0.354 means two samples are the same plant, or as close as identical twins.

Then we added a stricter rule. Chains can mislead: if A matches B and B matches C, A and C might only be cousins (the "my mate knows a guy who knows a guy" problem). So a family below only counts if every pair in it passes.

A strain name tells you what the seller called the plant. The DNA tells you what the plant is.

Some families read like a who's who of clone-only legends, and fans of each would swear on their grandmother's grow tent that they're totally different. Girl Scout Cookies, Forum Cut, Candyland and Animal Cookies share a genome. So do Sour Diesel, NYCD, Chem and Headband. Grand Daddy Purple, Purple Urkle and Mendocino Purple measure as one plant too.

Then there's Headband, the double agent. Samples sold as Headband turn up in two different families: one with the OG Kush cuts and one with Sour Diesel. Both stories have circulated for years. The DNA says different growers are selling different plants under the same name.

And the other way round: one name, many plants#

If one plant can wear fifty name tags, one name can also hide several plants, like a stage name shared by a whole bunch of different acts. For seed-grown and autoflower lines that's expected; every seed is a fresh roll of the genetic dice. For names sold as a specific cut, it's a surprise.

Take fourteen samples sold as Blue Dream: they came from five different plants. Twelve sold as Girl Scout Cookies came from six. Of the 101 names in this panel with three or more samples, only 13 (about one in eight) were one plant every time, and 64 (nearly two in three) were three plants or more. Buying by name is a bit like opening a mystery box.

So are family trees fiction?#

No, and this is the part that surprised us most. We checked the stated family trees in CannaLib against measured kinship. A real parent and child share about half their DNA, a kinship of around 0.25. We cut the claims some slack: a pair "fits" if it reaches at least half of that. By that rule, a stated parent and child fit as close relatives 27.3% of the time. Two random strains fit 1.9% of the time. That's 14 times better than chance, like a lottery ticket that wins 14 times as often as a random one. Not a guarantee, but definitely not a coin toss.

So roughly one stated parent in four shows up in the DNA, against about one in fifty for random pairs. Lineages carry real signal. But most single claims couldn't be confirmed on the samples tested, and there are two honest reasons. A shop sample is a plant sold under a name, not the plant the breeder crossed. And some stated lineages are simply wrong.

We also found 398 stated relations where the "parent" measures as the same plant as the "child". Grand Daddy Purple and Purple Urkle. Hitman OG and OG Kush. Candyland and Platinum Girl Scout Cookies. Those aren't crosses. They're one cut that picked up a new name along the way, like a band that rebrands and hopes you forget the first album.

What this doesn't say#

It doesn't say which name is the "real" one, and it doesn't accuse anyone. A name collected at one shop on one day is a snapshot. Mislabels happen too, and a few samples sit in families where they plainly don't belong. What the DNA gives you is a tiebreaker for every "bro, that's definitely the same cut" argument, and a way to stop arguing from memory.

CannaLib strain cards already show measured nearest relatives for 862 strains. The family counts in this article come from our newest analysis and will appear on strain cards as "sold under N names" in an upcoming release.