Methodology

Sources & weighting

Every lineage claim on this site is a publicly published claim, recorded from a source that put it in public view, and cited to that source. Nothing here is private data, and nothing is presented as this site's own discovery. A strain's parentage is a fact about a plant, not anyone's proprietary content — what this site contributes is the work of collecting those claims, checking them against each other, and being explicit about which ones actually hold up.

Not all sources are equal, and most public cannabis datasets are more related to each other than they look. The large consumer aggregators and the datasets built on top of them trace back to a shared pool of folklore. Three sites agreeing is often one claim copied three times, not three independent confirmations. So sources are tiered by the kind of authority they can actually carry, and only some tiers can produce a confirmed rating on their own:

• Tier A — primary. A breeder's own account of a cross they made, published either on their own site or in an interview with a named publication. This is first-person testimony from the person who did the breeding, and it is authoritative alone. Reported testimony counts here on the same footing as a breeder's own catalogue: what matters is that the originator is speaking, not which medium carried it.
• Tier B — curated genetics catalogues and expert review. Archived breeder-attributed catalogue pages, and cannabis journalism by named authors with verifiable credentials and direct access to breeders. Both carry attribution that can be traced to a person or a house, which is what separates this tier from anonymous aggregator fields. The tier attaches to that byline, not to the publication carrying it — a site that runs some Tier-B-qualifying journalism is not itself "a Tier B site," and an unattributed or pseudonymous byline on the same site is recorded at Tier C instead, claim by claim.
• Tier C — consumer aggregators and reported coverage. Large public strain databases with structured lineage fields, and cannabis journalism that reports on a strain's genetics without quoting the originating breeder directly. This tier holds two quite different kinds of source — an edited publication with named editorial responsibility is not the same as an anonymous database field — but both are used the same way here: for coverage and cross-checking, never as standalone confirmation. The entire point of the tiering is that aggregator agreement alone cannot produce confirmed.
• Tier D — dispute signal. Unattributed community claims: forum threads, comment sections, and social posts with no identifiable source behind them. Never confirms anything. Used only to flag that a lineage is contested — which is genuinely useful, because contested lineage is often first visible in community discussion long before it reaches any catalogue.

One disclosure specific to Tier B: one of the sources in this tier is operated by this site's own editor. A claim from that source is not independent corroboration of the editor's own contributions to this site — it is recorded and counted here as a single voice, never as two agreeing sources' worth of evidence.

How claims are cited

Every individual claim is recorded with its source and its tier in the database behind this site. The confidence rating on a strain page — confirmed, disputed, or unverified — and the count of supporting sources behind it are both computed directly from that record, not decorative. Full per-source detail — which sources agreed, which dissented, and their exact claims — is being surfaced progressively rather than shown on every strain page today: it's already live on thefull list of disputed lineages and on a growing set of individual edges, with the goal of reaching every disputed strain directly.

Source selection, tier assignment, and review of contested cases are directed editorially by a reviewer with working knowledge of cannabis genetics — not by an automated process alone. Where the record cannot reliably distinguish a breeder from a retailer, or a genuine phenotype from a naming variant, those calls are made by hand and recorded.

Confidence ratings

Every parent-child edge on this site carries one of three ratings:

• confirmed — either a Tier A source, or three or more independently-run breeders (each counted individually, not collapsed into "Seedfinder said so") agreeing with no credible contradiction.
• disputed — real, credible disagreement between sources. A small amount of dissent doesn't flip an otherwise well-supported edge to disputed — dissent has to reach at least half the weight of the support behind it to count as a genuine split, not a stray outlier claim.
• unverified — everything that isn't well-corroborated: a single source, aggregator-only agreement, or no real cross-checking at all yet.

Contradiction always outranks agreement in this system. A correctly-labeled disputed edge is more useful than a falsely confirmed one — showing you where genetics genealogy is genuinely contested is the actual product here, not a clean-looking family tree. See the full list of disputed lineages for every case where sources disagree.

Vote counts & dissent

Wherever you see a count like "6 supporting / 1 dissenting" next to a strain, that's a direct tally of independent sources for that specific edge — not a popularity score, and not weighted by how enthusiastic a source sounds. Independent breeders are counted individually because that's where real corroboration comes from: five Seedfinder pages attributed to five different breeders is five data points, but five aggregator sites copying the same folklore is one. Dissent is recorded even when it doesn't change the final rating, so you can see exactly how contested (or not) an edge actually is, not just its final label.

How corrections are reviewed

Anyone can submit a correction — a wrong parent, a missing one, a duplicate strain, or anything else recorded incorrectly. Submissions are collected, not published, and never change anything on their own: a correction is a claim, not a source, and it's weighed the same way every other claim on this site is. A human reviews each one by hand against what's already on file before anything is applied.

Corroboration matters here the same way it does everywhere else in this methodology. Several independent submissions agreeing on the same fix — the same strain, the same issue, the same proposed correction — carry real weight, the way independent breeders agreeing on a lineage claim do; a single submission is a lead worth checking, not a verdict. Nothing is auto-applied even when submissions agree, but agreement is what moves something to the front of the review queue.

Flavor families

The nine color-coded families you see across the site (Kush/Earth, Haze/Citrus, Diesel/Gas, Cookies/Bakery, Gelato/Dessert, Skunk/Funk, Purple/Grape, Candy/Sweet, Landrace/Heritage) are aflavor and naming-genre classification, not a genetic one. They're assigned from a strain's name, its documented lineage, and its recorded aroma/flavor profile — never from a genetic test. Two strains in the same family can be entirely unrelated by descent; what they share is how the name or aroma reads. Where a flavor reading and documented lineage genuinely disagree, the flavor reading wins for classification purposes, because this taxonomy is explicitly about genre, not ancestry — ancestry is what the lineage graph itself is for.

Aroma profile

Many strain pages carry an aroma profile: a nine-point radar showing how the strain's smell and taste read across nine registers. It is built from Leafly's crowd-reported flavor data, not from lab measurement. Leafly scores 47 flavor terms from its users' reviews, and each score is an open-ended relative number where zero is roughly average. It is not a percentage and not a concentration. We group the 47 terms into nine registers by what they smell like, and score each register as the average of whichever of its terms the strain has on record. The grouping is deliberately not statistical: in Leafly's data, terms that appear together mostly track how loudly reviewers describe a strain, not what it smells like, and following that would put cheese with the fruit and diesel next to grape.

Every radar uses the same fixed scale, so any two strains' shapes can be compared directly. The ring marked typical is Leafly's zero: a point outside it means reviewers noted that register more than usual, and a point inside it less. The center is the bottom of the scale, not zero. A register with none of its terms on record is labeled no data and drawn with dashed edges down to the center. That means nothing was recorded, not that the smell is absent.

• 3,143 strain pages (47.6%) show a radar, which needs at least 4 of the nine registers on record.
• 326 (4.9%) have only one to three registers, too few to draw a meaningful shape, and show a one-line summary instead.
• 3,138 (47.5%) have no Leafly flavor data at all and show nothing: either Leafly has no flavor data for them, or they weren't in our Leafly snapshot.

The data is a single snapshot from our Leafly collection, not a live feed, and it won't change until that snapshot is refreshed.

The nine registers

The notes on chemistry below are general background on what typically drives each smell, not a measurement of any particular strain.

Citrus

Bright peel and juice: lemon, lime, orange, grapefruit. Often driven by limonene, the same compound that makes citrus rind smell like citrus rind.Leafly terms: citrus, lemon, lime, orange, grapefruit.

Candy Sweetness

Confection rather than fruit: sugar, tropical brightness, an almost artificial sweetness. Ester-driven, and the register modern dessert-lineage breeding has pushed hardest.Leafly terms: sweet, honey, mango, pineapple, tropical, apple, pear, peach, apricot, tree fruit.

Dark Fruit

Deep and jammy: grape, blueberry, plum, berry. Comes from esters rather than terpenes, which is why two strains can share a terpene profile and taste nothing alike.Leafly terms: grape, blueberry, berry, plum, strawberry.

Floral

Blossom and petal: lavender, rose, violet. Linalool sits behind much of this; it is what makes lavender smell like lavender, and it turns up in cannabis more often than people expect.Leafly terms: flowery, lavender, rose, violet.

Herbaceous

Green, leafy and cooling: sage, tea, mint, cracked pepper. Pinene and terpinolene drive the fresh green edge; the peppery bite is caryophyllene, the only terpene known to bind directly to a cannabinoid receptor.Leafly terms: sage, tea, spicy herbal, pepper, mint, menthol.

Earth

Soil, forest floor, bark and pine. Myrcene and humulene do most of the work here — humulene is the same compound that makes hops smell like hops.Leafly terms: earthy, woody, pine, tobacco.

Funk

Sour, savory, barnyard: skunk, aged cheese, pungency. Like Gas, this is sulfur chemistry rather than terpenes, which is why it survives drying and curing so stubbornly.Leafly terms: skunk, cheese, blue cheese, pungent.

Gas

Fuel and solvent: diesel, chemical sharpness, ammonia. Not a terpene smell at all — the gassiness in cannabis comes from volatile sulfur compounds, the same family that gives skunk and garlic their punch.Leafly terms: diesel, chemical, ammonia, tar.

Dessert & Cream

Bakery and dairy: vanilla, butter, coffee, toasted nut. The Cookies and Gelato register, and the newest of the nine in cannabis terms — it barely existed before the 2010s.Leafly terms: vanilla, butter, coffee, nutty, chestnut.

Limitations — read this before trusting any single lineage claim

Most cannabis lineage is genuinely undocumented, and this site shows that rather than hiding it:

• 9,599 parent-child relationships are recorded in total — 1,386 confirmed (14.4%), 1,444 disputed (15%), 6,769 unverified (70.5%).
• Of the edges with a parseable source count, 6,742 (70.2%) rest on exactly one source with nothing to corroborate against — there's a real difference between "unverified because no one has checked" and "unverified because sources genuinely conflict," and this site doesn't yet distinguish the two in the rating itself.
• Tier A coverage is limited and being actively expanded. An early check of breeder sources targeted heritage strains — the ones whose lineage is most disputed — and found little primary authority available for them, because strains of that era predate any living commercial entity that could serve as a first-person source. Modern strains with an identifiable originating breeder are a better fit, and that work is underway.
• Tier B's coverage of modern breeders is thinner than its coverage of established European catalogues. Several contemporary breeders are listed with only a fraction of the strains they have actually released, so a strain's absence from this site is not evidence it does not exist.
• Coverage itself has gaps — some strains have no lineage claim from either source yet, and appear on this site with an empty ancestor list rather than a guessed one.

The confirmed percentage can fall even while the data is getting better, and it's worth spelling out with a real example rather than leaving it to look like a regression. The thehighestcritic.com ingest (2026-09-18) raised confirmed edges by 334 in absolute terms — 452 of those were existing edges that a newly added independent source corroborated for the first time, pushing them from unverified to confirmed. But the confirmed percentage fell over that same ingest, from 17.0% to 16.7%, because the ingest also added 1,376 new strains, and a newly recorded strain starts with a single source by definition — that's what "newly recorded" means. Every new strain widens the unverified denominator before anything has had a chance to corroborate it. A falling confirmed percentage next to a rising confirmed count is the expected shape of a growing database, not a decline in quality.

None of this is smoothed over in the ratings you see elsewhere on the site. A confirmededge here means real, independent corroboration was found — not that the question is permanently closed.