The Lithuanian Research Centre for Agriculture and Forestry ran seven years of field trials with glyphosate on wheat, barley and oilseed rape. The result is unusually clean. Spraying before the crop emerged left nothing in the grain: no residues at all, at the label rate or at double it. Spraying before harvest left residues every time, at any dose and any timing.
Whether there is glyphosate in your cereal comes down to one decision made by an agronomist: when the sprayer went out.
Pre-harvest desiccation: where the residue comes from
Glyphosate is applied to a standing crop about two weeks before harvest to even out ripening and dry down weeds. The practice is legal, and the Lithuanian trial covered both versions of it: on-label at 14 to 10 days before harvest, and off-label at 4 to 2 days.
Both produced glyphosate and its metabolite AMPA in the grain. Neither reached the maximum residue level set by the European regulation. The practice works within the limit, and it cannot be done without leaving a trace.
One more detail from the same trial removes a common hope. Residues stayed at steady concentrations in stored grain for longer than a year. Time does not help here.
Bran against white flour
The most useful part of the work is where the residue ends up after milling. With pre-harvest application at the label rate, glyphosate concentrated in wheat bran and rapeseed meal. In white wheat flour and in cold-pressed oil, no residues were found.
| Processed fraction | Glyphosate residue |
|---|---|
| Wheat bran | most of it |
| Rapeseed meal | most of it |
| White wheat flour | none detected |
| Cold-pressed oil | none detected |
The reason is straightforward: glyphosate is water-soluble and stays in the seed coat rather than the starchy core. Which leaves an awkward conclusion for wholegrain enthusiasts. Wholemeal flour, bran and flaked cereals carry the residue; a white roll does not.
That is not an argument against bran. It is an argument for comparing risks and benefits in the same units instead of in adjectives.
What turns up in stores
The Swiss Federal Food Safety and Veterinary Office tested 243 market samples by liquid chromatography with tandem mass spectrometry, at a very low limit of quantification. Cereals and pulses were the main contributors to dietary intake, and among individual foods the authors singled out pasta.
An oddity sits in the detail: every sample of wine and fruit juice, and nearly every sample of honey, tested positive, though at very low levels.
The key fact about that survey comes from the authors themselves. Products were not picked at random, they were picked because high residues were suspected. Even so, not a single sample exceeded the maximum residue level.
What urine shows
The German KarMeN study approached it from the other end and measured people rather than products: 24-hour urine from 301 participants. In 66.5 percent of samples neither glyphosate nor AMPA was detected at all. Quantifiable concentrations appeared in 8.3 percent of participants, and the rest showed traces.
Calculated intake came out far below the acceptable daily intake. The food pattern matched: urinary glyphosate was significantly associated with eating pulses, and urinary AMPA with eating mushrooms.
An MRL is a legal limit, not a harm threshold
A 2021 review supplies the figure that gets quoted in every argument: regulatory surveys indicate 99 percent of glyphosate residues in food sit below European MRLs or US EPA tolerances, and urine-based estimates put exposure below 3 percent of the current European acceptable daily intake, which is 0.5 mg per kilogram of body weight.
Before using that number, know who supplies it. All four authors of the review work at Bayer Crop Science, the company that owns the Roundup brand. That is not a reason to dismiss the review, it is a reason to check its claims against independent sources, which is what the sections above do.
The most valuable line in that review cuts against simplification from both directions: an MRL is a limit on legal pesticide use, not a health figure. Health information comes from comparing total intake with the acceptable daily intake, and the two get confused constantly. The same confusion runs through arguments about the Dirty Dozen list.
The cancer dispute
In March 2015 the International Agency for Research on Cancer classified glyphosate as probably carcinogenic to humans. The EU assessment and the joint WHO and FAO evaluation did not confirm that conclusion.
The account of why they diverged was written by EFSA staff together with the German risk assessment institute, so it should be read as one side of the dispute. On their explanation there are two causes: the parties used different data sets, particularly long-term rodent studies, and different methods for weighing the available evidence.
The most recent count is a 2026 meta-analysis of ten primary datasets. For ever-exposure, the odds ratio for non-Hodgkin lymphoma came out at 1.11, a modest effect. For the highest exposure category the association was significant and held across every sensitivity analysis.
One qualifier makes that number usable. The meta-analysis is about people who work with the herbicide: applicators, farmers, laborers. Bread and porridge do not appear in it, and an odds ratio from the field does not transfer to a loaf.
The wheat and gut theory
The popular version runs like this: people stopped tolerating wheat because of glyphosate killing off good gut bacteria rather than because of gluten. Canadian researchers at the University of British Columbia went through that literature specifically.
A mechanism exists. Glyphosate blocks the shikimate pathway, which operates in plants and bacteria, and opportunistic pathogens are more resistant to it than commensals are. On that basis the authors call dysbiosis from food residues possible. Similar questions are asked of food emulsifiers, and there human trials already exist, collected in our piece on E numbers.
Then comes the sentence that usually disappears in retellings. Research on glyphosate’s effects on the microbiome suffers from numerous methodological weaknesses, and those limitations make it impossible to draw any definitive conclusion about health through changes in the microbiome.
Key points
- Timing decides it. Pre-emergence spraying leaves no residue in grain; pre-harvest spraying always leaves some, below the limit.
- Residues hold steady in stored grain for more than a year.
- Glyphosate concentrates in bran and meal. None was detected in white flour or cold-pressed oil.
- In a Swiss survey of 243 samples, cereals and pulses contributed most, with pasta named specifically. Nothing exceeded the limit, even in a deliberately targeted sample.
- In 66.5 percent of urine samples from 301 German participants, neither glyphosate nor AMPA was detectable.
- 99 percent of residues sit below MRLs and exposure is under 3 percent of the 0.5 mg/kg acceptable daily intake. That figure comes from a review written by Bayer staff.
- An MRL marks legal use, not harm. Health is judged by comparing total intake with the acceptable daily intake.
- IARC called glyphosate probably carcinogenic in 2015; the EU and WHO/FAO did not agree. A 2026 meta-analysis gives 1.11 for ever-exposure and a significant association at the highest exposure, but it covers occupational use.
- The wheat-intolerance-through-dysbiosis theory has a mechanism and no proof: the studies are too weak methodologically to support a conclusion.
Why glyphosate use grew alongside herbicide-tolerant crops, and how those are regulated in Ukraine now, is covered in our piece on the new GMO law. Why washing does nothing about what sits inside a product is in our look at the baking soda soak. What appears in grain and bread after the field, when they are stored badly, is covered in our piece on mold on food. The wider picture sits in our food safety hub.
