Not Every Plant Cut Into a Rooibos Harvest Is Rooibos
A Cape weed common across rooibos country looks enough like the crop to be missed at harvest, and carries a compound the tea plant does not make on its own. Here is what South African chemists traced it to, and what independent testing has actually found.
In 2013 Germany's Federal Institute for Risk Assessment, the BfR, reported the first survey to look for pyrrolizidine alkaloids, a family of natural plant toxins, across common herbal teas, and found them in fennel, chamomile, peppermint, and nettle tea. It also found them in rooibos, which made no obvious sense: Aspalathus linearis, the plant this whole industry is built on, had never been shown to produce the compound at all. A 2017 chemistry study settled the question. Rooibos itself makes none of the toxin. Every documented trace comes from a Cape weed that looks enough like rooibos, and blooms the same shade of yellow, to be cut and dried right along with the harvest.
Spray residue is the worry you most likely bring to a rooibos tin, and on that count the record is reassuring. A 2022 South African study tested 100 retail rooibos products, natural, flavoured, and herbal blends, for thirteen common pesticides. Only one sample carried any detectable trace at all, of two compounds, both well under the European Union's own safety ceilings. The authors' own conclusion was plain: on pesticides, the tea tested clean. The weed is where the real, documented problem sits.
A weed that flowers the same colour, in the same row
The 2017 study, by Ben-Erik van Wyk, Maria Stander, and Hanno Long, working out of the University of Johannesburg and the University of Stellenbosch, went looking for the source. In September 2014 they surveyed seven rooibos-growing districts, from Nieuwoudtville in the north to Piketberg in the south, photographing and collecting every weed growing among the tea bushes that might carry the compound.
One plant stood out. Senecio angustifolius, known locally as bitterbos, turned up in abundance in nearly every plantation the team visited. It is a perennial shrublet, usually about a metre tall (roughly 3 feet) though it ranges from 0.4 to 1.8 metres, with thin, often reddish, erect stems and thread-like needle leaves. Its flower heads are a plain, petal-less disc, no ray florets at all, in the same yellow as a rooibos bloom.
Growers in different valleys have their own names for the plant, and the names describe the same problem from different angles: bitterbos for the bitter taste it leaves on the hands, ghwanobos ("ghwano bush") in the Agter-Pakhuis area, where farmers say it only became troublesome once fertiliser use began, sprinkaanbos ("locust bush") near Ceres, and dunsiektebos ("thin sickness bush") in the Sutherland district, a name that records what it does to livestock that graze it. Van Wyk and his colleagues found it especially hard to control for reasons that have nothing to do with chemistry: it resists common herbicides, it germinates and grows year-round instead of on a season like most weeds, and, most simply, a worker moving quickly through a row of rooibos can cut it without ever noticing it was there.
How the chemists confirmed it
A resemblance is a theory. The team wanted proof, so they ran mass spectrometry on samples of bitterbos from every district and on batches of rooibos tea that had already tested positive for the toxin. The match was close: the two dominant compounds in bitterbos, senecionine N-oxide and retrorsine N-oxide, made up about 76 and 14 percent of its total alkaloid load. In the contaminated tea samples, the same two compounds made up about 59 and 13 percent. Spiking a clean tea sample with the two isolated compounds and rerunning the analysis confirmed the identification directly.
They also tested rooibos plants themselves, hand-picked from fields with no weeds nearby, leaves washed before drying to strip any surface dust or pollen. Every one of those careful samples came back negative. Aspalathus linearis does not make pyrrolizidine alkaloids. That much is settled.
Why a hand-picked plant is not always a clean one
What was not settled is what happens to a rooibos plant growing beside a heavy bitterbos infestation. Several of those plants, sampled with the same care as the weed-free ones, tested positive anyway, in fields at Agter-Pakhuis and Piketberg in particular. The explanation the researchers landed on is lateral transfer: soil collected from directly around bitterbos roots carried the same alkaloids, while soil from weed-free ground did not, and a separate study had already shown a related Senecio species leaching its toxins into soil and being absorbed by neighbouring plants' own roots. A rooibos bush can end up carrying trace amounts of a compound it never produced, simply for growing next to a plant that did, with the toxin arriving through the ground rather than the harvest sack.
The rule the European Union wrote with rooibos specifically in mind
Since 1 July 2022, the compound has had an actual ceiling. Under EU food-contaminant rules, most dried herbal infusions may carry no more than 200 micrograms of the combined alkaloids per kilogram. Rooibos does not sit in that general category. The regulation names it directly, in a separate class alongside anise, lemon balm, chamomile, thyme, peppermint, and lemon verbena, dried herbal infusions the rulemakers evidently judged prone to naturally higher background levels, set at double the general ceiling, 400 micrograms per kilogram. A dried infusion meant for infants and young children is held to a tighter 75 micrograms per kilogram, and a ready-to-drink liquid infusion for that age group to 1 microgram per kilogram. Rooibos has a long-standing reputation, noted in the same 2017 study, as a tea given to babies for colic; the stricter infant number exists for exactly that population.
The alkaloid only turns harmful after the liver's own enzymes convert it; the untouched form the plant carries is not itself toxic. The tumour findings behind the word carcinogenic come from animal studies at doses well above ordinary tea drinking, and no clinical link between the compound and human cancer at typical dietary exposure has been established. Regulators set a numeric ceiling rather than a ban: a limit meant to bound the exposure, not to pull rooibos, fennel, or chamomile off the shelf.
What retail testing actually turned up
Before that European ceiling existed, one retail snapshot showed how wide the spread could be. In April 2016 the German consumer publication ÖKO-TEST bought 21 rooibos products off the shelf and sent them for laboratory testing. About two-thirds came back above the German risk agency's own guidance value at the time, five of them by more than tenfold, and measured levels ranged as high as roughly two milligrams per kilogram, five times today's EU ceiling for rooibos specifically. Only three products passed with a clear recommendation. All three were organic-certified. No comparable public retest appears to have been published since the EU limit took effect in 2022, so whether that spread has narrowed under the new rule is not yet on the public record.
Rooibos has pest problems of its own worth managing without heavy spraying, and the industry has long leaned on the low-chemical habits that keep a plantation's natural pest predators nearby: tolerating a ground cover of weeds instead of clearing them with herbicide. The plantations that skip herbicide to protect those predators are the same ones that give bitterbos room to establish itself among the rows.
Van Wyk and his colleagues recommend not more chemicals but more careful eyes: workers trained to name bitterbos on sight, walking a field before the harvest cutters move through, with particular attention to the plants tangled in among the rooibos where a quick pass would miss them. The weed comes in with hand-harvesting, and a more careful hand-harvest is what keeps it out. What no amount of inspection changes is where the compound comes from, and it does not come from the red bush.