At Least One Marker Detected
Forty-six percent of honey consignments sampled at EU borders were flagged as suspicious. The flag fires if any one of four chemical markers turns up, and the markers are nowhere near equally worth believing.
The number is 46 percent, and it has been in circulation since March 2023, when the European Commission published the results of a coordinated action called From the Hives: 320 consignments of honey sampled at EU border control posts, 147 of them flagged as suspected of not complying with the Honey Directive. Headlines rounded it to half the honey in Europe being fake, which is wrong twice over — the samples were imports, and European production covers roughly 60 percent of what Europe eats — but 46 percent of imports is bad enough that the rounding barely changes the argument. What caught me, reading the underlying chemistry paper, is not whether the number is too high. It is what kind of number it is.
Here is the decision rule, in the words of the Joint Research Centre chemists who ran the analysis: a sample counted as suspicious “if at least one of the compounds analyzed was detected.” Four marker classes were in play. Mannose. Difructose anhydride III. 2-acetylfuran-3-glucopyranoside, or AFGP. And oligo- and polysaccharides with a degree of polymerisation between 6 and 11. Find any one of them and the consignment joins the 147.
The markers are not worth the same
AFGP, as far as the literature goes, does not occur in honey; it arrives with certain syrups, and if it is there, something was added. Mannose is a different animal, and the JRC authors say so themselves: “certain honeys, e.g., linden, eucalyptus, jujube, or chestnut, could naturally contain mannose,” and they recommend looking for further markers when it shows up. The high oligosaccharides are vaguer still — the paper concedes that “the origin of oligosaccharides and polysaccharides with a DP higher than 6 or 7 in authentic honey remains uncertain,” and that honeydew honeys may carry them natively. Mannose and the high-DP sugars were the two most frequent flags in the exercise.
Somebody has put numbers on that difference. A Defra working group chaired out of the UK’s Laboratory of the Government Chemist spent six meetings between October 2023 and December 2024 building a framework for interrogating the databases behind honey authenticity claims, and in it they work the markers through as likelihood ratios — the forensic device for saying how much more probable a finding is if a sample is adulterated than if it is genuine. AFGP, absent from two reference sets of 160 and 106 authentic honeys, comes out near 105: moderately strong support. Caramel colouring lands near 9. Mannose, reported in about 8 percent of 5,300 honeys in one study and 25 percent of Chinese honeys in another, falls to between 4 and 12 — limited support, in the language forensic scientists use when they are being careful.
So the markers differ in evidential weight by an order of magnitude, and the rule that produced the 46 percent treats them as the same event. The paper notes that two to five markers were present in 44 percent of the suspicious cases — which means the majority of the 147 rest on a single detection, and we are not told how those single detections break down by marker. I would like that table more than almost anything else in the dataset.
The China figure is where this bites hardest. Sixty-six of 89 Chinese consignments were flagged, 74 percent — and if a quarter of authentic Chinese honeys carry mannose, as one of the studies Defra cites found, then a mannose-only flag on a Chinese consignment is close to uninformative. I cannot tell from the published work how many of those 66 were mannose-only. Nobody outside the laboratory can.
The hatched bars are the part of the chart I keep returning to. Ukraine was the second-largest origin group, 74 consignments to China’s 89, and has no published rate. Nor do Argentina, Mexico or Brazil. Three origins out of twenty have numbers attached in public, so the 46 percent is an aggregate over a distribution nobody is shown.
Who collected the sample seems to matter
One figure in the paper I did not expect: among the member states doing the sampling, Romania returned the lowest suspicion rate, 3 of 15, and France the highest, 17 of 21. All the testing was central, at the JRC, so this is not laboratory-to-laboratory variation. It is either a real difference in what arrives at Romanian and French ports or a difference in what the two authorities chose to put in the bag. A fourfold spread between member states, under a headline that is a single percentage, is worth more than a footnote.
It does not help that the official documents disagree with each other about the basics. The Commission’s own page dates the sampling to November 2021 through February 2022; the chemistry paper says January 2021 to June 2022; the SANTE report’s timeline has it starting in October 2021. The same report records 320 samples against a target of 520 to 690, so the exercise delivered roughly half of what it planned. The Commission describes the samples as “randomly taken” while also stating that “the results were never expected to provide a representation of the EU market for honey” and “shall not be generalised or extrapolated to other situations.” Both sentences are defensible. Together they are the sound of an institution publishing a number it does not want held against it.
A test that cannot say how much syrup is in the jar, read against a database nobody outside the laboratory can see, is still evidence. It is just evidence of a strength somebody has to state out loud.
The reproducibility problem underneath
In 2022, chemists at the Laboratory of the Government Chemist published an autopsy of three commercial certificates of analysis behind a UK newspaper story about supermarket own-brand honey. The certificates returned adverse opinions in 20 of 21 categories; on the data the reviewers found “a much more nuanced picture” — markers reported without being named, foreign sugars reported without identification, caramel reported with no quantitative result, limits of detection missing. They also cite a beekeeper-led round of 14 samples sent to two providers and run through three techniques, where “no sample was returned as adulterated by all three techniques in both” providers.
The interpretive machinery sits in proprietary databases — commercial NMR libraries holding data on more than 20,000 samples, which the same paper notes do not necessarily span the range of global honeys, and which their owners had offered to open only in a dispute over formal enforcement. That is the barrier the Defra framework was built to lower: 150 standard questions about a database’s scope, representativity, metadata and validation, so an outside party can at least ask whether the library behind a verdict is fit for it. When the group trialled the questionnaire in the summer of 2024, two database holders expressed interest and one took part. Ninety-six percent of the questions proved relatively easy to answer — with a caution from the reviewers worth more than the percentage: “a lot of ‘no’ answers doesn’t mean the data doesn’t exist.”
What it added up to
Enforcement followed the arithmetic you would expect from evidence of this strength. Of 123 exporters controlled, 70 were flagged; of 96 EU importers, 63 had taken at least one suspicious consignment; 44 of those had been investigated by the time of the report, and 7 sanctioned. Consignments were turned back — 244 tonnes here, 20 tonnes there, a 35-tonne market ban — but 147 suspicious samples produced seven sanctions, and the report calls even that “a provisional insight.”
The law has moved on a slower clock. Directive 2024/1438 made country-of-origin labelling of blends mandatory — percentages, descending order, principal field of vision — and those rules have applied since 14 June 2026. It also obliges the Commission to adopt methods for detecting adulterated honey, but by 14 June 2028, with traceability rules by 2029 and a platform in the meantime to advise on whether to set up a Union reference laboratory. Until then member states are told to use “internationally recognised validated methods where possible,” a careful way of saying that the markers which generated the 46 percent are not yet legally the measure of anything.
Where I think this is weaker than it looks
The loudest attack on the study comes from the Global Honey Organization, an industry body in Lugano, which calls the report a scientific debacle and a tool of protectionism and wants it retracted. That is an interested party and the protectionism charge is their interpretation, not a finding. But their technical complaint — that mannose and high-DP oligosaccharides occur naturally and were used as if they did not — is the same complaint the JRC authors make about their own markers, more politely, in their own paper. An argument can be self-serving and correct at once.
My own position is thinner than I would like. I have the open-access chemistry paper, the Commission’s report, the Defra framework and its trial summary, and the 2022 reproducibility work; the per-marker counts, the method-agreement counts and the country table in the paper’s figures are either unpublished or locked inside images I cannot read. The likelihood ratios I lean on are worked examples built on stated assumptions, and an AFGP ratio of 105 derived from zero hits in 266 authentic honeys is bounded by those 266, not by nature.
What I am fairly confident of is the shape of the thing. Imported honey averaged €2.32 a kilo in 2021 while rice syrup ran €0.40 to €0.60, and a margin like that does not go unexploited; fraud is real and probably common. The measurement system cannot yet say how common, because its most frequent marker is its weakest and its reference libraries are private. Both of those are true, and 46 percent is the sound of them being collapsed into one number that neither supports.
Sources
- Detection of Honey Adulteration by LC–High-Resolution Mass Spectrometry: Results from an EU Coordinated Action — Separations
- EU coordinated action on honey: sampling, investigations and results — European Commission (DG SANTE)
- EU coordinated action: Honey (2021-2022) — European Commission, Food Safety
- Food fraud: how genuine is your honey? — EU Joint Research Centre
- Framework for interrogation of honey authenticity databases (FA0198) — Defra / Government Chemist
- FA0198 Annex 4: review exercise summary report — Defra / Government Chemist
- Honey authenticity: the opacity of analytical reports, part 1 — npj Science of Food
- Directive (EU) 2024/1438 amending the honey directive — EUR-Lex
- EU 'From the Hives' report, a scientific debacle — Global Honey Organization
- Honey authenticity and fraud, and what steps are being taken to improve testing — Food Manufacture
- What is the demand for honey on the European market? — CBI