# Honey: research notes

Date: 2026-10-19. Category slug `honey`, vertical formulations. Scope: packaged honey for eating (liquid, creamed, comb, chunk, Manuka/graded). Honey-flavoured syrups only as an adulteration note.

Quote labels: "verbatim" = text read from the page or from a PDF extracted locally with pdftotext. "fetched excerpt" = sentence returned by WebFetch, which summarises with a small model. "search summary" = a claim that appeared only in a search-result digest; treated as one weak source.

## Seed

Seed run: `seed.py --category honey` gave the olive-oil `harvest_date` example and the persona "a ship's cook, who thinks in shelf life, storage and what works without the ideal setup".

1. Honey's equivalent of harvest date. Honey does not go off, so the date itself matters less than for oil. The honey number that plays the same role is HMF (hydroxymethylfurfural), which starts near zero in the comb and rises with heat and time. It is the one figure that says whether a jar was cooked or stored hot, and it is almost never on a listing. Kept as facet `hmf_level`, researched, top-N backfill. Diastase activity is the companion marker; folded into the HMF note rather than a separate facet.
2. What the ship's cook notices first. Moisture content, because honey over about 18.6% water can ferment in a warm galley; crystallisation, because a jar that sets solid is useless unless you have a warm-water bath; and creamed honey, which never drips and never surprises you. Kept: `moisture_content` (researched), `crystallisation_tendency` (derived from floral source and processing), `form` (creamed as a storage-stable form). Dropped: container type (squeeze vs jar), which a photo answers.

## Sources and what each supports

### Codex Alimentarius, Codex Standard for Honey, CODEX STAN 12-1981 (rev. 2001)
URL: https://www.fao.org/input/download/standards/310/cxs_012e.pdf (PDF read locally with pdftotext; verbatim)
- Definition: "Honey is the natural sweet substance produced by honey bees from the nectar of plants or from secretions of living parts of plants or excretions of plant sucking insects on the living parts of plants, which the bees collect, transform by combining with specific substances of their own, deposit, dehydrate, store and leave in the honey comb to ripen and mature."
- "2.1.1 Blossom Honey or Nectar Honey is the honey which comes from nectars of plants. 2.1.2 Honeydew Honey is the honey which comes mainly from excretions of plant sucking insects (Hemiptera) on the living parts of plants or secretions of living parts of plants."
- "No pollen or constituent particular to honey may be removed except where this is unavoidable in the removal of foreign inorganic or organic matter."
- "3.3 Chemical or biochemical treatments shall not be used to influence honey crystallisation."
- Moisture: "(a) Honeys not listed below - not more than 20% (b) Heather honey (Calluna) - not more than 23%".
- Sugars: fructose+glucose "not less than 60 g/100g" (honeydew 45); sucrose "not more than 5 g/100g" with listed exceptions at 10 and 15.
- Diastase: "in general not less than 8 Schade units and in the case of honeys with a low natural enzyme content not less than 3 Schade Units."
- HMF: "shall not be more than 40 mg/kg. However, in the case of honey of declared origin from countries or regions with tropical ambient temperatures, and blends of these honeys, the HMF content shall not be more than 80 mg/kg."
- Conductivity: blossom "not more than 0.8 mS/cm"; "Honeydew and chestnut honey and blends of these - not less than 0.8 mS/cm"; exceptions include Manuka.
- Labelling: "6.1.6 Honey may be designated according to floral or plant source if it comes wholly or mainly from that particular source"; "6.1.12 Honey which has been filtered in such a way as to result in the significant removal of pollen shall be designated filtered honey."
Confidence: high. Supports mechanism, moisture, HMF, diastase, conductivity, honeydew family, filtered-honey labelling.

### UMF Honey Association, "Overview" of the UMF grading system
URL: https://www.umf.org.nz/umf-comparison/overview/ (fetched excerpt, table returned as printed)
- Table: UMF 5+ MGO 83, leptosperin >100, DHA 150, HMF <40; UMF 10+ MGO 261, leptosperin >150, DHA 250; UMF 15+ MGO 512, leptosperin >200, DHA 400; UMF 20+ MGO 826, DHA 500; UMF 25+ MGO 1197, DHA 500; HMF <40 at every grade.
- "MGO is one of four factors measured to ensure consumers around the world get a pure potent genuine New Zealand mānuka honey product."
- "NPA is a single compound measurement whereas the UMF™ rating is made up of multiple factors essential for validating the true quality of a batch of New Zealand mānuka honey."
Disagreement: search digests and several brand charts (New Zealand Honey Co, Manuka Hut) give 83 / 263 / 514 / 829 / 1200; Comvita's own UMF 10+ jar says "MGO 263+". The official page today prints 261 / 512 / 826 / 1197. Difference is immaterial for buyers; we use the official page and note the older numbers.
Confidence: high.

### MPI (New Zealand Ministry for Primary Industries) mānuka honey export definition, via The True Honey Co blog and Manuka Doctor NZ
URLs: https://truehoney.co.nz/blogs/the-true-honey-co-blog/test-for-the-best-mpi-releases-its-scientific-definition-of-manuka-honey ; https://www.manukadoctor.co.nz/blogs/content/how-to-read-a-manuka-honey-label (fetched excerpts). MPI's own page (https://www.mpi.govt.nz/export/food/honey-bee-products-requirements/manuka-honey-science-definition/) returned empty to the fetch tool; the two secondary pages agree.
- Monofloral: "3-Phenyllactic acid – at least 400mg per kg", "2'-Methoxyacetophenone – at least 5mg per kg", "2-Methoxybenzoic acid – at least 1mg per kg", "4-Hydroxyphenyllactic acid – at least 1mg per kg", "DNA from Mānuka pollen (level required < Cq 36, approx. 3 fg/µL)".
- Multifloral: 3-phenyllactic acid 20 to 400 mg/kg (Manuka Doctor; also a search digest of the MPI consultation document).
- "From February 2018, Mānuka honey for export will be tested to ensure it meets the MPI definition."
Confidence: high on the thresholds (two sources agree, both quoting MPI).

### FSAI (Food Safety Authority of Ireland), Honey legislation page, on Directive (EU) 2024/1438
URL: https://www.fsai.ie/enforcement-and-legislation/legislation/food-legislation/honey (fetched excerpt)
- "all countries of origin must be indicated on the label in the principal field of vision and listed in descending order of weight, including their representing percentage of weight in the total product."
- "A tolerance of 5 % shall be allowed for each individual share within the blend, calculated on the basis of the operator's traceability documentation."
- Packs under 30 g: country names may be replaced by ISO alpha-2 codes.
- "will apply from 14 June 2026."
- The directive "omits 'filtered honey' from the list of permitted definitions"; pollen "shall not be considered an ingredient".
Confidence: high.

### Eurofins, "Origin labelling for honey"
URL: https://www.eurofins.com/de-de/food-analysis/food-news/food-testing-news/origin-labelling-for-honey/ (fetched excerpt)
- Countries of origin of blends "in descending order, with the percentage of each country's origin"; "blend of honey from EU countries and non-EU countries" no longer acceptable; 5% tolerance; 30 g code rule; transposition by 14 December 2025, application from 14 June 2026.
- Eurofins verifies declared proportions with "microscopic pollen analysis".
Confidence: high. Second source for the origin rule. A search digest of the directive text adds that ultrafiltration means a mesh under 100 µm and that ultrafiltered honey "should no longer be authorized to be marketed and labeled as honey"; used only as a judgement note.

### FDA, Guidance for Industry: Proper Labeling of Honey and Honey Products (Feb 2018)
URL: https://www.fda.gov/files/food/published/PDF---Guidance-for-Industry--Proper-Labeling-of-Honey-and-Honey-Products.pdf (PDF read locally; verbatim)
- "If a food contains only honey, the food must be named 'honey'... The common or usual name may also include the source of the honey, such as 'Clover Honey'".
- "You do not have to declare the floral source of honey on the label. However, you may label the honey with the name of the plant or blossom if you or the honey producer has information to support the conclusion that the plant or blossom designated on the label is the chief floral source of the honey."
- "A product consisting of honey and a sweetener cannot be labeled with the common or usual name 'honey'"; must be e.g. "Blend of honey and corn syrup".
- No standard of identity: the Congressional Research Service brief (below) states "To date, FDA has not established a standard of identity for honey." and that "In 2011, FDA denied the 2006 petition".
Confidence: high. Supports floral_source being self-declared in the US, and the blend-naming rule.

### Congressional Research Service, "Ongoing Efforts to Address Fraud and Adulteration of Honey" (IF12185)
URL: https://www.everycrsreport.com/reports/IF12185.html (fetched excerpt)
- 2021 import shares: "India 124.7 [million lbs.] 26%; Vietnam 123.5 [million lbs.] 25%; Argentina 95.3 [million lbs.] 20%; Brazil 76.0 [million lbs.] 16%".
- Transshipment: "Honey from China that is transshipped through another Asian country and falsely sold as honey from that second country—usually to avoid higher customs duties and tariffs."
- Adulteration: "cheaper sweeteners such as corn syrup, rice syrup, sugar beet syrup, or cane sugar."
- "To date, FDA has not established a standard of identity for honey."
Confidence: high. Supports country_of_origin and authenticity facets.

### JRC (European Commission Joint Research Centre), "Food fraud: how genuine is your honey?" (23 Mar 2023)
URL: https://joint-research-centre.ec.europa.eu/jrc-news-and-updates/food-fraud-how-genuine-your-honey-2023-03-23_en (fetched excerpt)
- "A total of 320 honey consignments - imported from 20 countries - were randomly sampled between November 2021 and February 2022."
- "147 samples (46%) were suspicious to be adulterated, which means non-compliant with the general provision of the EU Honey Directive."
- "Sugar syrups made from maize are now rarely used to extend honey. They have been replaced by syrups made mostly from rice, wheat or sugar beet."
Disagreement: the Global Honey Organization (industry body) calls the 46% figure a "scientific debacle" and notes only 14% were confirmed non-compliant in the 2015-17 action (search digest). The Eurofins chapter below gives the 2015-17 numbers directly.
Confidence: high on the numbers; medium on what share is truly adulterated.

### Raezke, K.-P., "Honey", chapter in Food Integrity Handbook (Eurofins), 2018
URL: https://cdnmedia.eurofins.com/european-west/media/12153881/10_chapter-honey_final.pdf (PDF read locally; verbatim)
- Moisture: "Mature honey typically contains 13 to 23 % water [20]. If it is over 18 %, there is a risk of fermentation, roughly related to the level of contamination by yeasts and the water content."
- EU low-enzyme rule: diastase "not less than 3" for "honeys with low natural enzyme content (e.g. citrus honeys) and an HMF content of not more than 15 mg/kg".
- Resin technology: "bringing moisture level of honey up to 40 % and then reducing it to 18-19 %... the process can remove pollen and thus disguise country of origin or floral source. It also removes certain colour components of the honey, transforming dark honey into lighter, more acceptable product."
- Acacia: "The characteristic on which the acacia premium is based is the fact that it remains liquid for a long time due to its high fructose/glucose ratio." "Acacia is considered 'pure acacia' if the pollen consists of 20 % or more Robinia pollen." "Rapeseed honey crystallises very readily."
- Pollen: "in some cases the specific pollen may be 'under-represented' such as for citrus and lavender, whereas for others, such as forget-me-not, the pollen is 'over-represented'." "Pollen is also available as a product and the determined adulterator could filter out all the pollen and add back pollen of choice."
- HMF: "Excess heating during processing or unsuitable storage conditions can [raise] HMF content, making it a useful indicator of honey quality."
- NMR: "Since it is independent from potential manipulations of pollen, it also provides a complementary tool to check the declared botanical and geographical origin".
- 2015-17 EU control plan: "2 264 honey samples were collected... More than 10 % of the honey samples checked by EA/LC-IRMS did not conform to published benchmark purity criteria... Around 20 % of honey either declared as blends of EU honeys, or unblended honeys bearing a geographical reference... were suspected to contain added" sugars.
Confidence: high.

### Food Safety News, "Tests Show Most Store Honey Isn't Honey" (Nov 2011)
URL: https://www.foodsafetynews.com/2011/11/tests-show-most-store-honey-isnt-honey/ (fetched excerpt)
- "76 percent of samples bought at groceries had all the pollen removed"; "100 percent of the honey sampled from drugstores... had no pollen"; "77 percent of the honey sampled from big box stores... had the pollen filtered out"; "more than 60 jars... in 10 states"; analysed by "Vaughn Bryant, a professor at Texas A&M University".
- "The FDA has sent a letter to industry stating that the FDA does not consider 'ultra-filtered' honey to be honey".
- "without pollen there is no way to determine whether the honey came from legitimate and safe sources".
Confidence: medium (one investigation, 2011, US). Supports the processing facet's pollen note.

### National Honey Board, "Definition of Honey and Honey Products" (approved 15 June 1996)
URL: http://www.ccnybees.org/uploads/1/0/5/8/10585924/honeydefs.pdf (PDF read locally; verbatim)
- "Raw Honey: Honey as it exists in the beehive or as obtained by extraction, settling or straining without adding heat." "Commercially Raw Honey: Honey as obtained by minimum processing. This product is often labeled as raw honey."
- "Strained Honey: Honey which has been passed through a mesh material to remove particulate material (pieces of wax, propolis, other defects) without removing pollen." "Filtered Honey: Honey processed by filtration to remove extraneous solids and pollen grains."
- "Whipped Honey: Honey processed, by controlled crystallization, to a smooth spreadable consistency. Also called 'Cremed Honey,' 'Spun© Honey,' 'Whipped Honey,' 'Churned Honey,' 'Candied Honey' or 'Honey Fondant.'"
- "Crystallized Honey: Honey in which part of the natural glucose content has spontaneously crystallized from solution as the monohydrate."
- "Comb honey: Honey presented in its original comb or portions thereof." Extracted honey forms: "(1) liquid, (2) crystallized or granulated, or (3) partially crystallized".
- "Ultrafiltered Sweetener Derived from Honey: Honey from which all materials not passing a specified submicron membrane pore size have been removed. Materials removed include most proteins, enzymes and polypeptides."
Confidence: high. Supports form and processing values and vocabulary.

### Honey Bee Suite (Rusty Burlew), "The truth about raw honey"
URL: https://www.honeybeesuite.com/raw-honey-means-buyers-sellers/ (fetched excerpt)
- NHB definition quoted; "carries no legal weight". FDA sentence quoted: "Raw honey is honey which was not filtered or heated above normal ambient temperature."
- Raw does not make honey safe for infants: pasteurisation does not kill botulism spores.
Confidence: high. Supports the judgement that "raw" is undefined in law.

### Honey Bee Suite, "Don't be confused: whipped honey is crystallized, not whipped"
URL: https://www.honeybeesuite.com/monday-night-myth-creamed-honey-is-whipped-honey/ (fetched excerpt)
- "Creamed honey is controlled granulation of honey which results in extremely small sugar crystals"; "Creamed honey is not whipped honey. There shouldn't be any air in your creamed honey" (both quoting Betterbee).
Confidence: high.

### Honey Bee Suite, "Why did my honey crystallize and then ferment?"
URL: https://www.honeybeesuite.com/why-did-my-honey-crystallize-and-then-ferment/ (fetched excerpt)
- "Honey needs to be about 18.6 percent water or less for long-term storage." When glucose crystallises "the liquid portion of the honey is more than 18.6 percent water... the honey may ferment."
- Pasteurisation "heating to 145°F for 30 minutes" kills yeast but damages honey.
Confidence: high.

### Hamdan, K., "Crystallization of Honey" (beekeeping primer, Apeldoorn)
URL: https://oygardsbi.com/onewebmedia/XXX_2BHoney_Crystallization.pdf (PDF read locally; verbatim)
- "Generally, the fructose ranges from 30- 44 % and glucose from 25- 40 %... What crystallizes is the glucose, due to its lower solubility."
- "Honey high in glucose sugar, with a low fructose to glucose ratio will crystallize more rapidly, such as alfalfa, cotton, dandelion, mesquite, mustard and rape... Honey with a higher fructose to glucose ratio (containing less than 30% glucose) crystallizes quite slowly and can stay liquid for several years without special treatment, for example, robinia (black locust), sage, longan, tupelo and jujube/sidr".
- "Raw honey (unheated and unfiltered) contains bits of wax, pollen and propolis, and crystallizes faster. Honey that has been processed (e.g. heated and filtered) will remain in its liquid form longer than raw honey".
- "Honey crystallization is most rapid around 10-15 ºC (50- 59 ºF)... Honey resists crystallization best at higher temperatures more than 25 ºC (77 ºF). When the temperature is 40 ºC (104 ºF) the crystals dissolve. Temperature above 40 ºC (104 ºF) will damage the properties of honey."
- Speed table: Very slow: acacia (Australian), black locust, cranberry, litchi, longan, milk vetch, milkweed, sage, sidr, tulip poplar, tupelo. Slow: bell heather, blackberry, borage, buckwheat, chestnut, citrus, eucalyptus, fireweed, linden, maple, hawthorn, rosemary, sourwood, Spanish lavender, thyme. Rapid: alfalfa, apple/pear/plum/cherry, clover, cotton, dandelion, common lavender, phacelia, field bean, goldenrod, holly, ivy, mesquite, mustard, oilseed rape, raspberry, star thistle, sunflower, wild thyme.
- Re-liquefy "at 35-40 ºC (95-104 ºF). The temperature should not go beyond 40 ºC (104 ºF)". US packers "heat liquid honey to 63 ºC (145 ºF) for thirty minutes, or 71 ºC (160 ºF) for one minute".
- Storage: optimum "21 to 27 ºC (70-80 ºF)"; "Don't store in the refrigerator."
Confidence: high for the mechanism and the floral table (matches the NHB sheet below); medium for any single floral source.

### National Honey Board fact sheet "Honey Crystallization" (hosted by BJCP)
URL: https://legacy.bjcp.org/mead/crystal.pdf (PDF read locally; verbatim)
- "Honey with a low glucose-to-water ratio is likely to remain liquid, avoiding crystallization."
- Table 1 (adapted from White et al. 1962, USDA Tech. Bulletin 1261): higher-than-average crystallisation tendency (+): alfalfa, cotton, dandelion, mesquite, prune, rape; lower (-): acacia, cranberry, gallberry, grape, Mexican clover, milkweed, palmetto, raspberry, sage, sourwood, sunflower, tupelo.
- "The Dyce process is often used to make cremed honey. The method involves adding starter nuclei to honey after it has been heated twice [to 120 °F (49 °C) and 150 °F (66 °C)] and then strained."
- Storage: "Moderate temperatures [50–70 °F (10–21 °C)] generally encourage crystallization."
Disagreement with Hamdan: raspberry and sunflower are "rapid" in Hamdan and "lower than average" in the NHB/White table. We list only sources the two agree on in the facet values.
Confidence: high.

### Smiley Honey, "What Kinds of Honey Crystallize Fastest and Why?"
URL: https://www.smileyhoney.com/blogs/honey-school/kinds-of-honey-crystallize-fast (fetched excerpt; producer blog)
- Clover "typically solidifying within 2 to 6 months" (G:F 0.85); canola "within 2 to 4 months"; acacia "glucose-to-fructose ratio of 0.61... can remain perfectly liquid for over two years"; tupelo "remaining liquid for 1 to 2+ years".
Confidence: medium (producer blog, but consistent with Hamdan and IHC).

### Persano Oddo, L. & Piro, R. (2004), "Main European unifloral honeys: descriptive sheets", Apidologie 35 S38-S81 (International Honey Commission)
URL: https://www.ihc-platform.net/descriptivesheets.pdf (PDF read locally; verbatim)
- Method: "all three complementary approaches have to be taken into account, melissopalynological, sensory and physico-chemical. Melissopalynological analysis does not classify unequivocally unifloral honeys, and sensory evaluation may suffer from a certain subjectivity."
- Robinia (acacia): "Robinia pollen is under-represented"; sold "liquid (due to the high fructose content) and very light coloured and flavoured"; average colour 12.9 mm Pfund; F/G ratio 1.61.
- Rapeseed (Brassica): "characterised by quick granulation, due to the high glucose content"; "some European laboratories requires a F/G ratio lower than 1"; colour 26.2 mm Pfund; F/G 0.95; "Crystallisation rate: quick".
- Chestnut (Castanea): "Castanea pollen is strongly over-represented, and many laboratories require a percentage of at least 90%"; colour 87.9 mm Pfund; "Crystallisation rate: slow"; sensory "Bitterness: medium... Persistence/aftertaste: long".
- Heather (Calluna): "the pollen percent is often under the 45% limit"; "Crystallisation rate: moderate"; typical thixotropy (gel consistency).
- Citrus: colour 15.0 mm Pfund; F/G 1.24.
Confidence: high. Supports monofloral verification, the colour-to-source mapping and crystallisation values.

### Viper's bugloss honey paper (PMC5627913), on the 45% rule
URL: https://pmc.ncbi.nlm.nih.gov/articles/PMC5627913/ (fetched excerpt)
- "viper's bugloss honey is considered monofloral when Echium pollen represents at least 45% of total pollen load", citing Louveaux, Maurizio & Vorwohl (1978) "Methods of melissopalynology", Bee World 59(4).
Confidence: high for the 45% convention. A search digest of a 2026 Foods paper adds the over-represented 60 to >90% and under-represented <5 to 30% ranges; treated as corroboration only.

### True Source Honey, FAQs
URL: https://www.truesourcehoney.com/faqs (fetched excerpt)
- "The True Source Certified® seal requires honey to be fully traceable, from hive to table, tested to confirm authenticity and audited by a third party."
- Methods named: "EA/LC-IRMS Testing", "NMR or HRMS testing".
Confidence: high.

### Sweetree Honey (NZ producer), "What is HMF?"
URL: https://www.sweetreehoney.co.nz/blog/post/45795/What-is-HMF/ (fetched excerpt)
- "In the hive honey would normally be below 1 mg/kg"; "levels soon start to rise with ambient temperatures above 20°C"; "It is usual for HMF to be below 10 mg/kg in fresh extracted honey."
Confidence: medium (producer page). Second source for "fresh is under 10": search digest of rawhoneyguide and kashmiril ("typically below 10 mg/kg in fresh extracted honey", "most batches test below 5 mg/kg"). Codex gives the 40/80 limits.

### Honest Beekeeper, "What is the proper moisture content of honey?"
URL: https://honestbeekeeper.com/what-is-the-proper-moisture-content-of-honey/ (fetched excerpt)
- "Honey with more than 18% moisture can ferment over time"; "International guidelines agree that high-quality honey should have a moisture content of below 20%"; bees cap cells "around 18%".
Confidence: medium.

### Goodfriend Honey glossary, "Honey fermentation"
URL: https://goodfriendhoney.com/glossary/honey-fermentation.html (fetched excerpt)
- "Fermentation occurs when honey's moisture content exceeds approximately 18-20%"; "any honey reading above 18.6% moisture is at elevated risk"; yeasts are "species of Zygosaccharomyces and Torulaspora".
Confidence: medium. Together with Honest Beekeeper, Eurofins and Honey Bee Suite this fixes the 18 to 18.6% risk line.

### Beesource forum thread, "18.25% Moisture Content Acceptable?" (DROPPED from references: never read, search summary only, and the URL now returns 402)
URL: https://www.beesource.com/threads/18-25-moisture-content-acceptable.319408/ (search summary only)
- Beekeepers: "If you know its at 17% moisture you can discard fermentation issues. If it is over 18% you can be suspicious"; "up to 18.6% moisture is still Grade A honey per the USDA".
Confidence: low as a quote. Not cited in category.json.

### Raw Honey Guide, "Honey Quality & USDA Grade Estimator" and "Honey Color Scale Chart"
URLs: https://rawhoneyguide.com/tools/honey-quality-grader ; https://rawhoneyguide.com/tools/honey-color-scale (fetched excerpts)
- Grades: "Grade A: ≥90 score, ≤18.6 % moisture; Grade B: ≥80, ≤18.6 %; Grade C: ≥70, ≤20.0 %; Substandard: <70". "USDA honey grading is a voluntary, fee-for-service program administered by USDA Agricultural Marketing Service (AMS)." "Grading is not legally required to sell honey."
- Colour: "Water White ≤8 mm | Extra White >8–17 mm | White >17–34 mm | Extra Light Amber >34–50 mm | Light Amber >50–85 mm | Amber >85–114 mm | Dark Amber >114 mm". "USDA color designation is separate from quality grade." Flavour examples per class are "illustrative, and their actual color can cross class boundaries."
Confidence: medium (secondary); the USDA standard PDF (ams.usda.gov) returned 403 to the fetch tool. Hive to Honey below confirms the scores and, on the page itself, all seven colour ranges (the fetch excerpt returned only four; a blind check read the page and found all seven).

### Hive to Honey, "Honey Grades Explained"
URL: https://hivetohoney.com/honey-grades-explained/ (fetched excerpt)
- Grade A 90 points, B 80, C 70; "voluntary grade standard for extracted honey"; Pfund: Water White 8 mm or less; Extra White over 8–17; Light Amber over 50–85; Dark Amber over 114. "Dark Amber is a color class. Grade B is a quality grade. They are separate classifications." Canada uses No. 1/2/3; the EU "does not use the U.S.-style retail Grade A/B/C system".
- Search digest (Bee Inspired Goods, Netties Bees, Beekeepers Realm): "Honey does not require official inspection to carry a grade mark, and a packer can print 'Grade A' based on its own assessment"; "most honey on grocery store shelves has never been officially graded by the USDA".
Confidence: medium-high.

### UC Davis Honey and Pollination Center, Honey Flavor and Aroma Wheel (via UC ANR Bug Squad blog)
URL: https://ucanr.edu/blog/bug-squad/article/taste-honey-5 (fetched excerpt)
- "The Honey and Pollination Center produced the Honey Flavor Wheel after six months of research and development"; "a group of 20 people--trained tasters, beekeepers and food enthusiasts--who worked together with a sensory scientist to come up with almost 100 descriptors"; categories "fruity, floral, herbaceous, woody, spicy, nutty, confectionary, caramel and earthy" plus "animal".
- Search digest (honey.ucdavis.edu): 99 descriptors, developed 2013-14 by Amina Harris.
Confidence: high that the wheel exists and what it shows.

### Foods 2024, "Comparison of Antioxidant Properties and Color of Selected Polish Honeys and Manuka Honey" (DOI 10.3390/foods13172666)
URL: https://doi.org/10.3390/foods13172666 (abstract via Semantic Scholar API; verbatim abstract)
- "Dark honeys, including buckwheat honey, showed increased phenolic content and superior antioxidant properties. The study revealed remarkable correlations between phenolic content, antioxidant capacity and color. Buckwheat honey showed higher antioxidant properties compared to manuka honey".
Confidence: high for the correlation; one study for the manuka comparison.

### Foods 2021, "The Relationship between the Browning Index, Total Phenolics, Color, and Antioxidant Activity of Polish-Originated Honey Samples" (DOI 10.3390/foods10050967)
URL: https://doi.org/10.3390/foods10050967 (abstract via Semantic Scholar API; verbatim abstract)
- TPC order "buckwheat > heather > acacia > multiflorous > linden > rapeseed"; "The variables of browning and TPC, ACW, and FRAP were positively correlated with each other."
Confidence: high. Second source for colour tracking phenolics.

### CDC, "Foods and Drinks to Avoid or Limit" (infant nutrition)
URL: https://www.cdc.gov/infant-toddler-nutrition/foods-and-drinks/foods-and-drinks-to-avoid-or-limit.html (fetched excerpt)
- "Honey given to children younger than 12 months may cause a severe food poisoning called botulism." "Do not give your child honey before 12 months."
Confidence: high. Category constant, not a facet.

### Thorne (UK beekeeping supplier), "Advice on Honey Labelling"; Pembrokeshire County Council Trading Standards, "Labelling of honey"
URLs: https://www.thorne.co.uk/honey-labelling-advice ; https://www.pembrokeshire.gov.uk/trading-standards-information/labelling-of-honey (fetched excerpts)
- Mandatory UK label: name, producer, "country or countries of origin", "best-before date", lot, weight. "Honey that is intended to have a shelf life in excess of three months is required to be marked with a best before date."
- Permitted names: blossom/nectar, honeydew, comb, chunk, drained, extracted, pressed, filtered, baker's honey; pressed honey may use "heat of up to 45 C"; baker's honey "may have been overheated"; "unmarked honey is assumed unfiltered".
Confidence: high. Supports the point that a best-before date on honey is a legal formality, and the form vocabulary.

### NorCal Nectar, "Organic vs conventional honey"; PollenOps, "Organic honey certification"
URLs: https://norcalnectar.com/blog/organic-vs-conventional-honey ; https://pollenops.com/organic-honey-certification (fetched excerpts)
- EU organic: "3-kilometer (about 1.9 miles) radius of organic or wild forage around hives"; US certifiers "typically require that forage sources within 2-5 miles of each certified yard location are free of prohibited substances"; "USDA NOP doesn't have a specific honey production standard... Certifiers use the livestock standards as the closest applicable framework".
- "Most organic honey on U.S. shelves is imported"; Brazil "the largest exporter of certified organic honey to the US"; "about 70% of honey consumed in the US is imported".
Confidence: medium (two secondary sources agree; a search digest of Organic Produce Network and the OTA says the same).

### FoodCrumbles, "Why honey crystallizes and how to decrystallize honey" (and other decrystallising guides)
URL: https://foodcrumbles.com/why-honey-crystallizes/ (search summary)
- "At about 40°C (104°F) all the sugars will redissolve"; beekeeper guides: "never heat honey above 40°C if you want to keep its enzymes"; warm-water bath "35–43°C".
Confidence: medium as a quote; Hamdan gives the same 35-40°C verbatim.

### Retail listings (what data is present)
- Amazon clover honey listings (search digest of ten listings, e.g. https://www.amazon.com/Coxs-Honey-Unfiltered-Delicious-Product/dp/B07L458M32; direct fetches returned HTTP 500 twice, and a third attempt returned only the title and meta description, not the product-detail table): structured fields seen are Flavor (e.g. Clover), Item Form (Liquid), Specialty (e.g. Preservative Free), Honey Type, Country as Labeled, Diet Type (Kosher, Paleo), Unit Count, Item Weight, Container Type (Bottle/Jar/Jug). Claims live in titles: "100% Pure Raw Unfiltered", "Unpasteurized", "USA".
- Vitacost, Comvita UMF 10+ (https://www.vitacost.com/products/comvita-manuka-honey-umf-10-mgo-263-8-8-oz-250-g-88563; fetched excerpt): "Product of New Zealand"; attributes "Raw & Certified UMF™", "Non GMO Project Verified", "Glyphosate Free", "Kosher", "Unpasteurized"; "minimum 263 mg/kg of Methylglyoxal (MGO)"; "monofloral manuka honey"; nutrition per 21 g: 60 kcal, 15 g sugars; warning "Not suitable for children under 1 year of age"; "Store in a cool, dry place".
- Manuka Hut, Comvita MGO 263+ (https://manukahut.com/products/comvita-manuka-honey-mgo-263-umf-10; fetched excerpt): "UMF licence holder number 1019", sizes 8.8 oz / 17.6 oz, "raw monofloral Manuka honey", use-by dates printed.
- Nate's Honey Co product page (https://nateshoney.com/products/raw-unfiltered-honey; fetched excerpt): "gently warmed for pourability and unfiltered to keep the natural goodness, like pollen, intact"; "never pasteurized"; "Purity Guarantee"; sizes 12/16/32/44 oz; nutrition per 21 g: 70 kcal, 17 g sugars; floral source and country not stated on the page.
Takeaways: processing claims and Manuka numbers are in titles and bullets, floral source is a field on Amazon, country of origin is a field on Amazon and a label line in the EU/UK, moisture and HMF never appear, pollen status appears only as the word "unfiltered", nutrition is near-identical across jars.

## Numbers used in category.json and their two sources
- Moisture ≤20% (Codex verbatim; Eurofins verbatim). Fermentation risk above 18 to 18.6% (Eurofins verbatim; Honey Bee Suite verbatim; Goodfriend; Honest Beekeeper); the earlier "below 17% nothing grows" line was dropped because its only source was the unread Beesource digest. USDA Grade A/B ≤18.6%, C ≤20.0% (Raw Honey Guide; Hive to Honey scores; AMS search digest; the AMS standard itself was never read).
- HMF ≤40 mg/kg, 80 tropical (Codex verbatim; Eurofins verbatim). Fresh honey under 10 mg/kg (Sweetree; search digest of two primers): medium confidence.
- Diastase ≥8 Schade, ≥3 for low-enzyme honeys (Codex verbatim; Eurofins verbatim).
- Conductivity 0.8 mS/cm boundary (Codex verbatim; search digest of a review and Raw Honey Guide).
- UMF minimum MGO 83/261/512/826/1197 (UMFHA official page; brand charts give 263/514/829/1200).
- MPI monofloral 400 mg/kg 3-PLA, multifloral 20-400 (True Honey Co; Manuka Doctor).
- Pfund seven classes (Raw Honey Guide verbatim; Hive to Honey all seven on the page; search digest of two more pages).
- 45% pollen convention (Echium paper verbatim; IHC sheets refer to "the 45% limit"); exceptions acacia 20% (Eurofins verbatim), chestnut ≥90% (IHC verbatim), citrus/lavender under-represented (Eurofins verbatim; IHC).
- Crystallisation fastest 10-15°C (Hamdan verbatim; NHB sheet 10-21°C "encourage"); re-liquefy ≤40°C (Hamdan verbatim; FoodCrumbles digest); pasteurisation 63°C/145°F 30 min (Hamdan verbatim; Honey Bee Suite verbatim).
- Adulteration: 46% of 320 imported consignments suspicious 2021-22 (JRC verbatim; search digest of the Separations 2025 paper); >10% of 2 264 samples non-conforming 2015-17 and about 20% of EU-blend or named-origin honey suspected of added sugar (Eurofins verbatim; an earlier draft misread this as doubtful origin, corrected 2026-10-19); 76% of US grocery samples pollen-free (FSN; single source, so stated as one 2011 investigation).
- Origin rule applies 14 June 2026 (FSAI; Eurofins).
- Infant limit 12 months (CDC; Comvita label).

## Against the existing add-on schema

The add-on `formulations/food-beverages/add-ons/honey.jsonc` is empty. It composes with `food-beverages-standard.jsonc` (sizing, nutrition, allergens) and `formulations-standard.jsonc` (ingredients, high_signal_markdown, net quantity). Under the AGENTS.md rules (vertical anchor, quality currency, hard filter, sizing; everything else is Jev markdown), here is where each facet of this document would land. Proposals only.

| Facet | In add-on today | Proposed home | Reason |
|---|---|---|---|
| honey_type (blossom / honeydew / Manuka) | no | hard filter enum on the leaf | Manuka is a different price class and buyers filter on it literally; honeydew is a legal name on the label. |
| processing (raw-unpasteurised / pasteurised / unspecified) | no | hard filter enum | "Raw" is the single most-searched honey word; absence of the claim is itself the signal. Pollen status (strained vs filtered) rides along in the same enum or as a boolean. |
| floral_source | no | hard filter (string or enum of the top 20 sources plus wildflower/blend) | Buyers filter on "clover" or "acacia" literally; Amazon already carries it as Flavor / Honey Type. |
| color_class (Pfund) | no | none in the schema; image pipeline | Rule 1: a photo answers it. The document uses it as the novice's flavour question, which the UI can derive from the image embedding or from the label words "light amber". |
| form (liquid / creamed / comb / chunk) | no | hard filter enum | Creamed and comb are bought by name; also visually obvious, but the enum is cheap and the title always carries it. |
| crystallisation_tendency | no | derived at query time from floral_source and processing | Not a stored field. |
| mgo_level (mg/kg) | no | quality currency, numeric | The number the whole Manuka sub-catalog is compared on; dense on every Manuka listing; worth augmenting from the UMF table when only UMF is given. |
| umf_grade | no | quality currency, ordinal (5, 10, 15, 20, 25) or fold into high_signal_markdown | Already named as an example grade in formulations-standard's markdown comment ("UMF 15+"). If mgo_level is the stored currency, UMF can stay in markdown. |
| country_of_origin | partly: zamana-standard has `made_in` as one ISO 3166-1 alpha-2 code | keep `made_in` for single-origin jars; blends need a multi-valued origin list with optional percentages (EU labels carry them from June 2026) | A single code cannot hold "Argentina 60%, Brazil 40%", which is what EU jars will print; US jars carry "Product of". Fraud risk makes this the one provenance fact worth a literal filter. |
| authenticity_verification (True Source / NMR-tested / UMF-licensed / MPI-tested / none) | no | high_signal_markdown | Rule 5 kills certification blocks; the markdown comment already asks for "grades and certifications... third-party tested". Keep it there. |
| moisture_content (%) | no | quality currency only for the top N; otherwise markdown | Rarely stated; worth augmenting from producer COAs for the top of the catalog, not for every row. |
| hmf_level (mg/kg) | no | same as moisture | Same. |
| organic_certification | no | high_signal_markdown | Certification; Rule 5. |
| flavour_profile (wheel descriptors) | no | high_signal_markdown | Free text from tasting notes; the ranker reads markdown. |

Add-on fields the document considers noise: none, since the add-on is empty. Standard fields worth flagging: `nutrition` is nearly constant across honeys (60 to 70 kcal and 15 to 17 g sugars per 21 g on every jar checked), so it should not be used to rank within honey; `allergens` is rarely relevant (honey is not a major allergen) but harmless.

## Branches

Added 2026-10-19. A branch is a claim about how an expert narrows the choice, so each one points to the sentence in the register or in these notes it follows from. `color_class` has no branches: pale or dark, the next step is the same (judgement 3: "We use colour as the question and floral source as the filter"). Blossom honey, liquid and chunk are the default path and need none.

- `honey_type` = Manuka (graded) → ask `mgo_level`, `umf_grade`; skip `color_class`. From the Manuka family in the mechanism: "bought for its methylglyoxal content rather than its taste and priced by that number", and `honey_type` why_experts_care: "Manuka is graded and priced by its methylglyoxal (MGO) content, or by UMF". Colour is skipped because the colour split exists to ask about flavour strength ("Colour predicts flavour strength") and Manuka is not bought for taste.
- `honey_type` = honeydew (forest) honey → skip `color_class`. From `honey_type` why_experts_care: "Honeydew honey is darker, less sweet and higher in minerals." The pale-or-dark question is already answered.
- `processing` = raw, unpasteurised → ask `crystallisation_tendency`, then `floral_source`. From `crystallisation_tendency` why_experts_care: "Raw honey sets faster because the pollen and wax specks it keeps act as seeds", and the crystallisation trade-off: "If you want a honey that stays liquid without heating, you are choosing from the high-fructose sources: acacia, tupelo, sage and sidr."
- `processing` = heated or pasteurised → skip `crystallisation_tendency`. From the processing trade-off: "Choose heated and filtered and it pours for a year or more."
- `form` = creamed or whipped, or crystallised or set → skip `crystallisation_tendency`. From `crystallisation_tendency` backfill_note: "Creamed honey has already set, so mark it not applicable", and the vocabulary entry for crystallised honey: "the glucose has come out of solution as crystals".
- `form` = comb → skip `processing`. From `form` why_experts_care: "Comb honey is the least processed form, since it arrives sealed by the bees."

## Trimmed detail (2026-10-19)

Sentences cut from category.json to meet the validator's caps, with the field each came from. No claim, number, name or threshold in the register was changed; what is listed here is what no longer appears in the field itself.

- summary: "The nutrition panel is the same on every jar." (Survives as judgement 13 and in the add-on section above.)
- mechanism.plain: "phenolic content (the plant compounds behind colour and antioxidant activity)"; "The nutrition panel barely moves: both jars checked give 60 to 70 calories and 15 to 17 g of sugar per tablespoon." Also "with small amounts of acids, enzymes, pollen and flavour compounds carried over from the plant" shortened to "plus traces of acids, enzymes, pollen and plant flavour compounds".
- splits[honey_type].rationale: Manuka is "priced per gram far above table honey" (the per-gram qualifier was cut here; it stays in `honey_type` why_experts_care); "asking first removes it or isolates it" became "isolates it".
- splits[color_class].rationale: no fact cut; "how strong it tastes and how much of the plant compounds it carries" became "flavour strength and plant-compound content".
- facet honey_type.why_experts_care: "Codex, the international food standard"; the named conductivity exceptions "Manuka, heather, lime and eucalyptus" (kept in the mechanism families text and in the Codex reference); "it also lets the word honeydew appear next to the name."
- facet processing.why_experts_care: "A 2011 investigation found 76% of grocery-store honey in ten US states had no pollen at all." (Food Safety News 2011, under Sources.) "keeps it liquid on the shelf" became "keeps it liquid".
- facet floral_source.why_experts_care: rapeseed "sets within a couple of months"; "the FDA only asks the packer to have information to support the chief floral source"; "monofloral (single-source)"; the pollen exceptions "acacia counts as pure at 20% pollen because it sheds little, chestnut needs about 90% because it sheds a lot" (the numbers stay in judgement 12's note and in the IHC and Eurofins entries).
- facet crystallisation_tendency.why_experts_care: "Storage at 10 to 15 °C speeds setting, so a fridge is the wrong place." "a jar stood in water at 35 to 40 °C goes back to liquid, and heating past 40 °C damages it" became "a warm-water bath at 35 to 40 °C reverses it; past 40 °C damages it".
- facet mgo_level.why_experts_care: "Comvita sells its UMF 10+ (MGO 263+) jar as everyday strength."; "A high MGO honey tastes medicinal, so buyers who want it for toast are paying for potency they will not enjoy." (The medicinal point stays in the mgo_level trade-off.)
- facet umf_grade.why_experts_care: the licence fee is paid to "the UMF Honey Association"; "shows the honey was not overheated or stored too long" became "shows no overheating or long storage".
- facet country_of_origin.why_experts_care: "around 20% of jars labelled as EU blends or with a named origin were suspected of containing added sugar"; the 2026 rule "ends the old blend of EU and non-EU honeys line"; "US jars of imported honey carry a Product of line, and most organic honey on US shelves comes from Brazil." (Product of stays in listing_field_hint; Brazil in organic_certification.)
- facet authenticity_verification.why_experts_care: "NMR profiling compares the jar against a database of genuine honeys" (stays in vocabulary); "Ministry for Primary Industries" now written MPI; "the four-marker or five-marker definition was met" became "the marker definition was met"; "self-declared on most retail jars" lost "retail".
- facet moisture_content.why_experts_care: "Beekeepers read it with a refractometer, a handheld optical meter, before bottling".
- facet hmf_level.why_experts_care: "A pasteurised jar that then sat in a hot warehouse can pass the raw-looking label test and still be well over the limit."; "(the lab unit for diastase)" after Schade units; "naturally low-enzyme honeys"; "80 for honey from tropical countries" became "80 for tropical origin".
- judgements[0], [4], [5], [7], [11] claims: reworded only; no fact cut.
- judgements[7].note: "about 20% of honey sold as EU blends or under a named origin was suspected of added sugar" (also under Numbers above).
- judgements[11].note: acacia and chestnut thresholds differ "because of how much pollen each sheds"; the viper's bugloss paper "argues in the same breath that 45% is too low for its own honey and proposes 70%".
- references codex_stan_12 supports[4]: "not less than 0.8 mS/cm" and "not more than 0.8" became "at least" and "at most"; all four named exceptions kept.
- references eurofins_honey_chapter supports[7]: full wording was "2015 to 2017 EU plan: 2,264 samples, more than 10% failed sugar-purity criteria, and about 20% of honey labelled as EU blends or with a named origin was suspected of added sugar".
- references ihc_descriptive_sheets supports[1]: "fructose-to-glucose ratio" abbreviated to "F/G ratio"; "per source" dropped; all numbers kept.
- references echium_pollen_paper supports[2]: "the paper itself" became "the paper"; "the 45% figure" became "45%".
- references rawhoneyguide_grader supports[3]: "returned 403 and was not read" became "returned 403, unread".
- references amazon_clover_listings supports[1]: "reported by the search digest only" became ": search digest only"; all eight field names kept.

## Corrections log

### 2026-10-19
- Branches: 6 added across 3 splits (honey_type 2, processing 2, form 2); color_class has none. Listed under Branches above with the sentence each follows from.
- Questions: 12 question fields rewritten to at most 12 words with no preamble: the 4 split questions and 8 cold-start questions (honey_type, processing, color_class, form, floral_source, crystallisation_tendency, country_of_origin, organic_certification). Each split question is now the same text as its facet's cold_start_question.
- Fields trimmed to cap: 27 (summary, mechanism, 2 split rationales, 10 why_experts_care, 5 judgement claims, 2 judgement notes, 6 reference supports entries). Every cut sentence that carried a fact, number or source is recorded under Trimmed detail above.
- Validator clean; category.md and viewer.html regenerated.

### Correction, 2026-10-19 (blind check of branches)
- processing = raw: `floral_source` removed from `then_ask`. The checker graded the leg WEAK: the register says raw honey sets faster, not that a raw answer makes floral source the next question. Floral source stays a filter on every path.