# Coffee filters: research notes

Date: 2026-10-25. Scope: the separate filter medium a brewer holds, sold on its own: disposable paper filters in every format (drip-machine baskets, wedge cones #1 to #6, V60-style cones, flat-bottom wave filters, Chemex papers, AeroPress discs, percolator discs and wraps, paper inserts for refillable pods, empty drip bags) and reusable inserts that replace them (metal mesh, nylon mesh, cloth). Not the dripper, brewer or carafe itself, including drippers with a built-in metal mesh (left to the sibling `manual-coffee-tea-prep`, `supplies > kitchen > manual_coffee_tea_prep`), not refillable pod cups sold as brewer accessories, and not charcoal water filters for coffee machines (Amazon returns these for "coffee filters"; they filter water, not coffee). Where a reusable metal cone is sold as "dripper plus filter" in one body (the LHS and MISETTO cones on Amazon), it belongs to the sibling; a mesh insert that sits inside someone else's dripper or machine stays here.

Quote labels: **verbatim** = the sentence in quotation marks was in text I read myself (curl plus tag stripping on HTML, the Shopify `products.json` feed, the NCBI efetch abstract text, pdftotext on a downloaded PDF, or Wikipedia raw wikitext); **fetched excerpt** = the fetch tool summarised or lightly reworded the page; **search excerpt** = the claim came from a search-engine summary because the page itself was not read. A search-excerpt-only number does not go in category.json.

Session note: the shared WebSearch budget for the session (200 calls across concurrent agents) ran out after five searches in this run. Everything after that was read from known URLs with curl, the NCBI E-utilities and Shopify product feeds.

## Seed

Seed output (verbatim from `seed.py --category coffee-filters`):

- Example: bedsheets `weave`: "the expert split is often not the number everyone advertises. Thread count is the generic trap here. Weave is what changes how the sheet feels and sleeps."
- Persona: "A field biologist, who thinks in what survives abuse in the rain"

1. **The coffee-filter equivalent of thread count is thickness, the number brands lead with ("20-30% thicker paper than most of the competition", Chemex; "Thicker Paper traps impurities", Melitta).** The thing that actually changes the cup is how fast and how evenly water drains through the paper, and that is set by pore structure, density and creping more than by thickness. Gagné's caliper and drawdown tests show it: Cafec (207 µm) drained slowest at 5.79 mL/s while Hario unbleached (203 µm) drained fastest of the papers at 15.3 mL/s; and Cafec's own spec sheet sells its thinnest paper (TH-1, 0.15 mm, high density) as its slowest and its thickest (TH-3, 0.28 mm, low density) as its fastest. **Kept** as the `flow_rate` facet, with thickness explicitly demoted in a judgement. Also kept: the material split (paper, mesh, cloth), which is the weave-like choice that changes body versus clarity.
2. **The field biologist notices first what survives being wet, hot and reused, and what ends up in the soil.** Two things survive the data check. (a) Cloth is the one format that fails by staying wet: Gagné gave up on cloth because it is "a hassle to properly clean and re-use" (verbatim) and Coffee Bean Corral says cloth "needs careful cleaning because retained oils and moisture can affect later brews" (verbatim). Kept inside `filter_material` and a trade-off, not as its own facet. (b) "Compostable" on a box is not one thing: Melitta's white and brown cones are "Commercially Compostable. Facilities may not exist in your area" (verbatim), BPI certifies commercial and home compostability under different schemes (verbatim), and Hario's Meteor paper contains polylactic acid (verbatim). Kept as `compostability`. **Dropped**: a wet-strength or burst-resistance facet. Melitta and If You Care advertise "double crimp" and "never break" seams, but I found no test that measures tearing, so it stays here.

## Sources and what each supports

### Specialist measurement

**Coffee ad Astra (Jonathan Gagné), "An In-Depth Analysis of Coffee Filters", 2019 (curl, text read in full)**
https://coffeeadastra.com/2019/08/04/an-in-depth-analysis-of-coffee-filters-2/
- verbatim thicknesses: "Chemex unbleached: 167 ± 23 μm Chemex bleached: 210 ± 22 μm Hario unbleached: 203 ± 21 μm Hario tabbed: 206 ± 21 μm Cafec: 207 ± 21 μm Hario tabless: 242 ± 22 μm V60 cloth: 690 ± 22 μm Aeropress: 120 ± 22 μm".
- verbatim measured drawdown with 150 g water in a Hario Switch: "Cafec: 5.79 ± 0.03 mL/sec Hario tabless: 6.89 ± 0.05 mL/sec Hario tabbed: 11.03 ± 0.02 mL/sec Hario unbleached: 15.3 ± 0.1 mL/sec V60 Cloth: 18.1 ± 0.3 mL/sec Osaka metal: 67.8 ± 0.6 mL/sec Chemex bleached: 7.23 ± 0.02 mL/sec Chemex unbleached: 9.82 ± 0.02 mL/sec".
- verbatim taint test (TDS added to distilled water by an unrinsed filter): "Hario tabbed bleached: 0 ppm Hario tabless bleached: 0 ppm Cafec bleached: 1 ppm Hario unbleached: 5 ppm".
- verbatim: "it seems clear that 150g of hot rinsing water was enough to deplete the filter of all dissolvable materials"; "It's therefore preferable to rinse unbleached filters with hot water"; "This is not true of any bleached filters".
- verbatim: "bleaching seems to deteriorate the quality of pore distributions"; unbleached filters came out best on uniformity and clogging indices.
- verbatim on metal: "the pores are typically so large that the flow is only constrained by the grind size"; "metal filters won't produce very clear brews like other pour over filters, instead they will produce a beverage with suspended solids and fines, with less clarity and more body."
- verbatim on cloth: "a hassle to properly clean and re-use".
- verbatim on AeroPress: "you can press harder, and make up for that difference."
- verbatim on idealised versus measured flow: "it seems that filters made of different materials or with different creping behave differently"; and "I observed no clear correlation between filter thickness and how offset the idealized flow rate was from the real flow rate".
- verbatim weights: "5.4 g for Chemex and 1.4 g for Hario".
- Supports: flow_rate, the thickness judgement, the bleaching taint and rinse judgement, the metal and cloth trade-offs, the AeroPress branch, the flow versus clogging grammar. Confidence high for the measurements; one tester, one batch, 2019, so the brand ranking may have moved.

### Academic (abstracts read verbatim via NCBI E-utilities efetch)

**Urgert R, Katan MB. The cholesterol-raising factor from coffee beans. Annu Rev Nutr 1997;17:305-24. PMID 9240930**
https://pubmed.ncbi.nlm.nih.gov/9240930/
- verbatim: "The responsible factors are the diterpene lipids cafestol and kahweol, which make up about 1% (wt:wt) of coffee beans. Diterpenes are extracted by hot water but are retained by a paper filter. This explains why filtered coffee does not affect cholesterol, whereas Scandinavian "boiled," cafetiere, and Turkish coffees do."
- Supports: the mechanism and `cholesterol_oils_filtered`. Confidence high (review by the group that identified the compounds).

**Orrje E et al. Cafestol and kahweol concentrations in workplace machine coffee compared with conventional brewing methods. Nutr Metab Cardiovasc Dis 2025;35(8):103933. PMID 40089392**
https://pubmed.ncbi.nlm.nih.gov/40089392/
- verbatim: "12 (4-24) mg/L and 8 (3-19) mg/l for home-brewed, paper-filtered coffees (n = 5). Boiled coffee had high concentrations of cafestol and kahweol, 939 mg/L and 678 mg/L, but having it poured through a fabric filter reduced the concentrations to 28 and 21 mg/L. Other coffee brews (percolator, French press) contained intermediate levels of cafestol (∼90 mg/L)".
- Units are as printed in the abstract; I use only the ratios between methods (same units throughout) and none of the absolute figures in the JSON.
- Supports: paper-filtered coffee lowest of the methods measured; cloth (fabric) removes much but not as much as paper. Confidence high for the ordering.

**Naidoo N et al. Cholesterol-raising diterpenes in types of coffee commonly consumed in Singapore, Indonesia and India. Nutr J 2011;10:48. PMID 21569629**
https://pubmed.ncbi.nlm.nih.gov/21569629/
- verbatim: "The traditional 'sock' method of coffee preparation used in Singapore resulted in cafestol concentrations comparable to European paper drip filtered coffee (mean 0.09±SD 0.064 mg/cup)"; "Similarly low amounts of cafestol were found in Indian 'filter' coffee that used a metal mesh filter (0.05±0.05 mg/cup). Coffee samples from Indonesia using the 'sock' method (0.85±0.41 mg/cup) or a metal mesh filter (0.98 mg/cup) contained higher amounts of cafestol comparable to espresso coffee."
- Supports: the disagreement in the diterpene judgement: cloth and metal mesh vary with how the coffee is made, and some metal-mesh coffee is as low as paper. Confidence medium (27 samples from shops).

**Rendón MY, Dos Santos Scholz MB, Bragagnolo N. Physical characteristics of the paper filter and low cafestol content filter coffee brews. Food Res Int 2018;108:280-285. PMID 29735059**
https://pubmed.ncbi.nlm.nih.gov/29735059/
- verbatim: "The highest cafestol and kahweol concentrations were obtained using paper filters with micro perforations, considering similar sized paper filters. The paper filters showed high fat permeability and grammages between 50.46 and 67.48 g/m2. The diterpene retention capacities of the filters produced in the different countries were similar. The results showed that the porosity of the paper filter and the particle size of the ground roasted coffee were determinant factors".
- Supports: `perforated` and its judgement. Confidence medium (one study, one coffee).

### Primers

**NutritionFacts.org, "Paper-Filtered Coffee and Cholesterol" (curl)**
https://nutritionfacts.org/blog/paper-filtered-coffee-and-cholesterol/
- verbatim: "if you make drip coffee with a metal mesh filter common in many machines and do not add a paper filter in the cradle, it would presumably be just as bad as French press coffee."
- verbatim: "most of the cholesterol-raising cafestol was retained by the coffee grounds, rather than actually getting stuck in the paper filter itself. In other words, "the principal function" of the paper filter is not necessarily blocking the compound itself, but blocking any fine particles that are carrying the compound."
- verbatim: "you can cut out more than 90 percent of cafestol by switching from a French press or coffee maker with a metal mesh filter to one with a paper filter."
- Supports: gold-tone machine baskets in `cholesterol_oils_filtered`; the fines mechanism. Confidence medium (secondary, but quotes the primary studies). The "more than 90 percent" figure is a single secondary source and stays out of the JSON.

**Wikipedia, "Coffee filter" (raw wikitext)**
https://en.wikipedia.org/wiki/Coffee_filter
- verbatim: "Paper filters remove oily components called [[diterpenes]] (like [[cafestol]] and [[kahweol]]). Metal, nylon or porcelain mesh filters do not remove these components."
- verbatim: "Common in the US are fan- or boat-shaped filters "#0" (similar to "100"), "#1" (similar to "101"), "#2" (similar to "102"), "#4" (similar to "1×4"), and "#6" (similar to "1×6") ... as well as basket-shaped filters in an 8–12 cup home size and larger restaurant sizes."
- verbatim: "Hario has cone-shaped paper filter bag sizes "01" (for 1 cup), "02" (for 1–4 cups) and "03" (for 1–6 cups)."
- verbatim: Chemex "A half-moon shaped filter paper (bleached: FP-2, unbleached: FP-2N) is used for the 3-cup holders ... The larger holders for 5 ..., 6 ..., 8 ..., 10 ... and 13 cups ... can alternatively use prefolded square sheets ... prefolded circle filters ... or unfolded circle filters".
- verbatim: cloth "allows more of the oil to come through than paper filters".
- verbatim: "Some [[baristas]] claim that paper filters exhibit a "paperish" taste and recommend washing out the filter with a flush of hot water".
- Note: Wikipedia gives "about 100 g/m2" paper; Rendón measured 50.46 to 67.48 g/m2. Disagreement; no grammage figure goes in the JSON.
- Supports: size naming (US # versus European 1×), the mesh and diterpene claim (second source), Chemex sizes (second source). Confidence medium.

**Wikipedia, "Bleaching of wood pulp" (raw wikitext)**
https://en.wikipedia.org/wiki/Bleaching_of_wood_pulp
- verbatim: "Chlorine dioxide is sometimes used in combination with chlorine, but it is used alone in ECF (elemental-chlorine-free) bleaching sequences."
- verbatim: "ECF bleaching can substantially reduce but not eliminate chlorinated organic compounds, including dioxins, from effluent." "TCF bleaching, by removing chlorine from the process, reduces chlorinated organic compounds to background levels in pulp-mill effluent."
- Supports: the ECF and TCF definitions in `whitening_method` and vocabulary. Confidence high for definitions.

**Coffee Bean Corral, "Coffee Filter Sizes Explained" (curl)**
https://coffeebeancorral.com/blogs/coffee-knowledge/coffee-filter-sizes
- verbatim: "Coffee filter numbers are size codes for cone-shaped filters, not cup counts. A #2 filter commonly fits 2–6 cup electric cone brewers, while a #4 commonly fits 8–12 cup cone coffee makers."
- verbatim: "An 8–12 cup brewer with a flat-bottom basket needs a basket filter instead, so confirm the holder shape before buying #4 filters."; "Cone holders narrow toward the bottom, while basket-style holders have a broad, flat base."
- verbatim: "Paper usually retains more oils and fine sediment, which produces a cleaner cup. Metal allows more oils and fine particles through, so the brew often has more body. Cloth falls between the two, but it needs careful cleaning because retained oils and moisture can affect later brews."
- verbatim: "Disc filters are small, round filters used in brewers designed around a flat circular format, including AeroPress-style systems."
- Supports: size mapping (second source after Melitta), the holder-shape rule in the brewer split, the clarity-to-body scale, AeroPress discs. Confidence medium (retailer blog that cites Melitta).

### Spec sheets (brand pages and Shopify product feeds)

**Melitta USA shop (products.json feed, read verbatim; product pages resolve)**
https://shoponline.melitta.com/products/4-cone-filter-paper-white-100-count
- verbatim: "Melitta #4 Cone Coffee Filters fit all 8-12 Cup Cone Coffeemakers"; "Melitta #1 Cone Coffee Filters fit all 1 Cup Electric Cone Coffeemakers"; "Melitta #2 Size Cone Filters fit all 2-6 Cup Electric & 1 Cup Non-Electronic Cone and Manual Coffeemakers"; "Melitta #6 Cone Coffee Filters fit all 10 Cup Non-Electric Cone Coffeemakers"; basket: "Fits 8 -12 cup basket-style electric coffee makers", "Fits 4 - 6 cup basket-style electric coffeemakers".
- verbatim: "Microfine Flavor Enhancing® Perforations allow the full coffee flavor to filter through".
- verbatim: "Chlorine Free**" with the footnote "** Elemental Chlorine Free".
- verbatim: "*Melitta Cone Coffee Filters are Commercially Compostable. Facilities may not exist in your area."; "BPI® Certified (Biodegradable Products Institute®) compostable."
- verbatim: "Thicker Paper traps impurities for a smooth, clean cup."; bamboo cones "made with 40% Bamboo pulp".
- Supports: size values, perforated, whitening_method (ECF), compostability. Confidence high for what Melitta states.

**CHEMEX Bonded Filters, Pre-folded Squares (curl)**
https://www.chemexcoffeemaker.com/products/chemex-bonded-filters-pre-folded-squares
- verbatim: "20-30% thicker paper than most of the competition."; "Oxygen cleansed using a bleach-free process"; "compatible with Six, Eight, and Ten Cup Classic CHEMEX® coffeemakers, as well as Five, Eight, and Thirteen Cup Handblown"; "Fits all CHEMEX® Coffeemakers except for the Three cup Class Series (CM-1C), Three cup Handblown Series (CM-1), Three cup Glass Handle Series".
- verbatim FAQ on the same page: "The white filter paper goes through an oxygen cleansing bath followed by a series of hot water baths. The brown filter paper goes through only the hot water baths"; "The FP-2 and FP-2N filters are a "halfmoon" in order to fit into the 15 ounce brewers."; slow brews: "The coffee is ground too fine and is clogging the filter."
- Supports: Chemex sizes, oxygen whitening, the thickness disagreement. Confidence high for brand statements.

**Hario USA, V60 paper filters 01/02/03, Meteor and Flow (products.json, verbatim)**
https://www.hario-usa.com/products/copy-of-paper-filter-for-01-drippers
https://www.hario-usa.com/products/v60-meteor-filters-02
https://www.hario-usa.com/products/flow-paper-filters-50ct
- verbatim: "These 01 filters have a capacity of 1-2 cups."; "These 02 filters have a capacity of 1-4 cups."; "These 03 filters have a capacity of 1-6 cups."
- verbatim Meteor: "With nearly 3 times the permeation rate of the classic V60 filters"; "Made from paper, abaca, and polylactic acid (PLA)".
- verbatim Flow: "Fluted, flat bottom ... Filter type-155".
- Note: Hario says 01 is 1 to 2 cups where Wikipedia says 1 cup; the JSON names sizes only, no cup counts for V60.
- Supports: V60 size values, wave 155 as a shared format, the fast-paper end of flow_rate, PLA in compostability. Confidence high for brand statements.

**Hario Europe, V60 Filter Paper 02 (products.json, verbatim)**
https://www.hario-europe.com/products/v60-filter-paper-02-size-box-100-pcs
- verbatim: "This paper, either natural or oxygen bleached (and therefore chemical-free)"; "NB : Before brewing your coffee, make sure to rinse the filter with boiling water".
- Supports: oxygen whitening (second brand), the rinse advice (second source with Gagné). Confidence high.

**CAFEC USA, TH-1 Light Roast and TH-3 Medium Roast papers (products.json, verbatim)**
https://www.cafecusa.com/products/copy-of-cafec-cup-4-light-roast-paper-filter-v60-02-lc4-100w
- verbatim: "1) Slower Flow Rate for suggested light roast ... 2) Medium Flow Rate for suggested dark roast ... 3) Faster Flow Rate for suggested medium roast"; "Thickness-1: Thickness 0.15 mm *Density: High"; TH-3: "Thickness-3: Thickness 0.28 mm Density: Low Two-Side Crepe ... let water flow smoothest".
- Supports: flow_rate as a property brands now sell on, and thickness not predicting flow (second source with Gagné). Confidence high for what Cafec states.

**If You Care, No 4 Coffee Filters (products.json, verbatim)**
https://www.ifyoucare.com/products/no-4-coffee-filters
- verbatim: "made from unbleached totally chlorine-free (TCF) paper".
- Note: an unbleached paper is chlorine-free by definition, so "TCF" on a brown filter says nothing extra. Amazon title: "If You Care # 4 Cone Shaped Unbleached All Natural Compostable Coffee Filters, 100 Count (Pack of 1), Chlorine Free".
- Supports: the vocabulary note on TCF. Confidence high.

### Standards and regulators

**US eCFR, 21 CFR 176.170, Components of paper and paperboard in contact with aqueous and fatty foods (eCFR versioner API, verbatim)**
https://www.ecfr.gov/current/title-21/chapter-I/subchapter-B/part-176/subpart-B/section-176.170
- verbatim: "Substances identified in this section may be safely used as components of the uncoated or coated food-contact surface of paper and paperboard intended for use in producing, manufacturing, packaging, processing, preparing, treating ... aqueous and fatty foods".
- verbatim: the finished food-contact surface, extracted under the intended conditions, "shall yield net chloroform-soluble extractives ... not to exceed 0.5 milligram per" (square inch; the sentence continues past my grep line).
- Supports: the food-contact judgement. Confidence high.

**BfR (German Federal Institute for Risk Assessment), "Regulations on papers & boards in Germany", S. Merkel, EFSA FIP Network, 2018-07-11 (PDF, pdftotext)**
https://www.efsa.europa.eu/sites/default/files/event/180710-p-19.3.pdf
- verbatim: "XXXVI/1 Cooking Papers, Hot Filter Papers and Filter Layers (hot filter papers, tea bags, boil-in-bag packages etc.)"; "Recommendation XXXVI/1 - Hot water extract according to DIN EN 647"; manufacturers are "responsible to comply to food regulations (especially Regulation (EU) No. 1935/2004)."
- Supports: the food-contact judgement (EU side). Confidence high.

**SGS SafeGuardS 133/24, "Germany's BfR Amends Recommendations ..." (2024-08-21, curl)**
https://www.sgs.com/en/news/2024/08/safeguards-13324-germanys-bfr-amends-recommendations-for-food-contact-material-and-articles
- verbatim: "Recommendation XXXVI/1 'Cooking paper, hot filter paper and filter layer'"; amended in 2024 to expand "the scope of silicates or mixed silicates to include sodium".
- Supports: XXXVI/1 is current (second source). Confidence high.

**BPI, Certification (curl)**
https://bpiworld.org/certification
- verbatim: "Commercial Only Certification certifies that items will break down in commercial compost facilities, while Commercial & Home Certification certifies that items will break down in both commercial compost facilities and in properly managed home compost bins"; "BPI uses ASTM D6400 as the base standard"; "BPI's Home Compostability Certification Program is based on the French Standard NF T 51-800".
- Supports: compostability values. Confidence high.

### Retail listings (field presence only)

**Amazon.com search results, 2026-10-25 (curl; three queries; ~150 titles read)**
https://www.amazon.com/s?k=coffee+filters , https://www.amazon.com/s?k=pour+over+coffee+filters , https://www.amazon.com/s?k=reusable+coffee+filter
- verbatim titles, for example: "Amazon Basics Basket Coffee Filters for 8-12 Cup Coffee Makers, White, 200 Count"; "Melitta #4 Unbleached Coffee Filters, Natural Brown Cone Filters, 1-Pack"; "Hario V60 Paper Coffee Filter, Size 02, White, 100ct Box"; "Kalita Wave Paper Coffee Filters I Larger Size 185 I 100 Count"; "Chemex Unbleached, Pour Over FSU-100, 100ct"; "GOLDTONE Reusable 8-12 Cup Basket Coffee Filter fits Black and Decker Makers"; "Reusable Coffee Filter 8-12 Cup Nylon Mesh Basket for Mr Coffee 2 Pack"; "TGP Organic Hemp Cloth Coffee Filter Cone No. 4"; "Pour Over Coffee Dripper, Stainless Steel Coffee Filter 600 Mesh Ultra-Fine"; "3 X Kalita: Wave Series Wave Filter 185 (2-4 Persons)".
- Field presence, counted by eye across the three pages: format or brewer, size code, material for reusables, and count appear in most titles. (Corrected 2026-10-25 after the blind check: colour and count are often missing; see the recount below.) Perforation, whitening process, flow and compost certification almost never appear; "Compostable" appears on a few (If You Care).
- Supports: `source: listing` for brewer_fit, filter_size, filter_material, bleaching and the pack_count dependency. Confidence medium: search-result titles, not full listing pages.

**Walmart, Melitta #4 listing (curl)**
https://www.walmart.com/ip/seort/8467273
- verbatim title: "#4 White Cone Coffee Filters, 100 Ct, Unbleached Paper" (a self-contradicting title: "White ... Unbleached"). Supports: colour in titles can be wrong, so the bleaching backfill note says to check the photo.

### Could not reach

Kalita official sites (kalita.co.jp no response, kalita-usa.com empty feed and no Wayback snapshot), Prima Coffee Kalita page (404), AeroPress paper filter product (not in the Shopify feed), Melitta Canada and Melitta global FAQ (404). WebSearch budget exhausted mid-run (see top), so the EPA's 1980s dioxin-in-paper work and a second measured flow study were not found; the dioxin history is not claimed in the JSON beyond the ECF/TCF definitions.

## Data points used for derived metrics

- `cost_per_filter` = price / pack_count. pack_count is the supplies-standard `sizing.pack_count`. Most titles give a count ("100ct", "200 Count"; "Pack of 2" multipliers need multiplying out), but not all: the blind check found 9 of 51 paper titles with no per-pack count, and my 2026-10-25 recount found about 10 of 47 (Melitta "1-Pack", "3-Pack", "6-Pack" titles, Katy's, Ninja pod, Presto, Mr. Coffee 5 cup). Those need the bullets.
- `reusable_breakeven_brews` = reusable_price / cost_per_filter of the paper the buyer would otherwise use. No sourced lifespan for a mesh or cloth filter, so the metric stops at break-even and makes no claim about how long the reusable lasts.
- No rinse-water metric: Gagné's 150 g hot water for one Hario unbleached 02 and his "about 4 times more" for Chemex (from the 5.4 g versus 1.4 g weights) is one tester and an inference; kept here only.

## Numbers in category.json and how many sources stand behind each

| Number or threshold | Sources |
|---|---|
| #2 fits 2 to 6 cup electric cones; #4 fits 8 to 12 cup cones | Melitta (verbatim), Coffee Bean Corral (verbatim) |
| 8 to 12 cup baskets | Melitta (verbatim), Coffee Bean Corral (verbatim), Amazon titles; Wikipedia "8–12 cup home size". The "4 to 6 cup basket" value was removed 2026-10-25 (Melitta only; Amazon shows 1-4, 4-5 and 2-5 cup baskets instead). |
| V60 01, 02, 03 size names | Hario USA, Wikipedia (cup counts differ; not used) |
| Chemex half-moon for 3-cup, squares and circles for 5 to 13 cup | Chemex (verbatim), Wikipedia (verbatim) |
| Wave 155 and 185 size names | 155: Hario Flow "Filter type-155" and Kalita USA "For use with a Kalita Wave 155 Dripper" (verbatim). 185: Kalita USA "For use with a Kalita Wave 185 Dripper" (verbatim) and Amazon Kalita 185 titles. kalitausa.com (not kalita-usa.com) answered on 2026-10-25. |
| Cafec 0.15 mm slowest, 0.28 mm fastest | Cafec only (brand spec); used with Gagné's measurement for the qualitative claim "thickness does not predict flow", no figure in the JSON |
| Chemex "20-30% thicker" | Chemex (verbatim), Wikipedia (repeats it, so not independent); Gagné's caliper disagrees. In the JSON only as Chemex's disputed claim, no measured figure |
| US #4 and #6 against European 1×4 and 1×6 | Wikipedia only ("similar to"); removed from the JSON 2026-10-25 for lack of a second source |

No other number appears in the JSON.

## Against the existing add-on schema

Add-on read: `supplies/kitchen/add-ons/coffee_filter.jsonc` (`compatible_with` string, `sizing.filter_size_standard` enum [no1, no2, no4, no6, basket, disc, cone_unsized, not_applicable, other, unspecified]) composed with kitchen-standard (`materials` with roles, `max_heat_temp`, `dishwasher_safe`) and supplies-standard (`high_signal_markdown`, `sizing.size_stated`, `sizing.pack_count`, dimensions, capacity, weight). Field kinds per AGENTS.md: vertical anchor, quality currency, hard filter, sizing. Proposals only; nothing edited.

| Facet in this register | In schema today | Proposed kind under AGENTS.md | Note |
|---|---|---|---|
| brewer_fit | partly: `filter_size_standard` mixes shape and size (basket, disc, cone_unsized beside no2, no4) | hard filter | Proposal: a `filter_format` enum [basket, wedge_cone, conical_v60, flat_bottom_wave, chemex, disc_aeropress, pod_insert, drip_bag, percolator] outside sizing (format does not change across size variants). `compatible_with` stays as the raw brewer string. |
| filter_size | partly: `filter_size_standard` | sizing (hard filter) | Proposal: keep the enum inside `sizing` but extend it to the formats buyers search by: [no1, no2, no4, no6, basket_4_6_cup, basket_8_12_cup, basket_commercial, v60_01, v60_02, v60_03, wave_155, wave_185, chemex_half_moon, chemex_6_13_cup, aeropress, other, unspecified]. Size varies per variant, so it belongs in sizing. `cone_unsized` and `disc` move to `filter_format`. |
| filter_material | yes, kitchen-standard `materials` (stainless_steel, fabric_textile, nylon, other; paper is missing from the enum) | vertical anchor | Add `paper` (and `abaca_paper` is not needed; fibre goes to markdown). Without `paper` the enum cannot record the commonest material in the leaf. |
| bleaching | no | hard filter | Proposal: `paper_color` enum [white, natural_brown, not_applicable]. In every Amazon title; cheap. |
| whitening_method | no | high_signal_markdown | Footnote-level ("Elemental Chlorine Free", "oxygen cleansed"); a sentence, not a column. |
| flow_rate | no | high_signal_markdown | Only Cafec states it as a product property; Hario Meteor states a relative claim. Not dense enough for currency. |
| cholesterol_oils_filtered | no | derived from material at query time | Not stored. |
| perforated | no | high_signal_markdown | Melitta-specific claim; a sentence. |
| compostability | no; AGENTS.md cuts certifications | high_signal_markdown | Commercial versus home is the useful distinction and is a certification, so it belongs in the markdown line, not a column. |
| (pack count) | yes, supplies-standard `sizing.pack_count` | sizing | Needed for cost per filter; keep. |

Add-on and standard fields this register considers noise for coffee filters: `max_heat_temp` (every filter meets boiling water by design; no listing states it), `dishwasher_safe` (matters only for the small reusable slice and is in reusable titles anyway; fine to keep null for paper), `sizing.dimensions` and `sizing.weight` (the size code replaces dimensions; nobody buys a #4 by millimetres), `sizing.capacity` (the cup count belongs to the brewer, and the size code already encodes it), `sizing.size_stated` duplicates the extended `filter_size_standard`.

## Branches

Each leg traces to a sentence in the register or above. The brewer split's `when` labels are its values; the question's answers map to values by the rule in the brewer_fit definition ("drip machine" goes to basket or wedge cone by holder shape; "other" goes by brewer name).

**Split 1, brewer_fit ("Drip machine basket or cone, V60, Chemex, AeroPress, or other?")**

- when `drip machine basket`, `wedge cone (#1 to #6)`, `flat-bottom wave`: then_ask filter_size.
  - Basket and wedge cone: "Coffee filter numbers are size codes for cone-shaped filters, not cup counts" and "confirm the holder shape before buying #4 filters" (Coffee Bean Corral); Melitta's basket pages "Fits 4 - 6 cup" and "Fits 8 -12 cup basket-style" (verbatim, on the basket product pages). In the register: filter_size why_experts_care.
  - Wave: 155 and 185 are different formats (Hario Flow "Filter type-155"; Kalita USA "For use with a Kalita Wave 155 Dripper" and "... 185 Dripper"; Amazon "Wave Filter 185 (2-4 Persons)").
- when `V60-style cone`, `Chemex`: then_ask filter_size, flow_rate. (Changed 2026-10-25: Chemex moved here from the first leg.)
  - filter_size, V60: Hario USA 02 page "These 02 filters have a capacity of 1-4 cups" and Hario Europe "We also have 01 size paper and 03 size paper available" (verbatim).
  - filter_size, Chemex: "Fits all CHEMEX® Coffeemakers except for the Three cup" (Chemex product page, verbatim) and "The FP-2 and FP-2N filters are a "halfmoon" in order to fit into the 15 ounce brewers" (Chemex FAQ, verbatim).
  - flow_rate: Gagné measured V60-fit papers (Cafec 5.79 to Hario unbleached 15.3 mL/s) and Chemex papers (bleached 7.23, unbleached 9.82 mL/s) in the same rig; Chemex FAQ: "The coffee is ground too fine and is clogging the filter" (verbatim). flow_rate is not novice-answerable and has no question, so this leg shows it as an annotation; the facet says "Shown, not asked." Wave is not on this leg: Gagné did not test a wave paper, and the backfill note now covers V60 and Chemex only.
- AeroPress disc: no branch (removed 2026-10-25). Gagné's "you can press harder, and make up for that difference" supports "matters less", not "irrelevant", and in the same paragraph he prefers Aesir papers for AeroPress on pore quality. The register says "flow matters less there" in flow_rate why_experts_care.
- No branch for `paper pod insert or drip bag` or `percolator disc or wrap`: nothing sourced changes downstream.

**Split 2, filter_material ("Paper you throw away, or reusable metal, nylon or cloth?")**

- when `metal mesh`, `nylon mesh`: then_ask cholesterol_oils_filtered; makes_irrelevant bleaching, whitening_method, perforated, flow_rate, compostability.
  - cholesterol_oils_filtered: "Metal, nylon or porcelain mesh filters do not remove these components" (Wikipedia, verbatim) and the NutritionFacts gold-tone line ("presumably"). The leg shows the annotation; it is not a question (the facet says "Shown, not asked"). Nylon rests on Wikipedia alone.
  - bleaching, whitening_method, perforated: defined for paper only (their definitions in the register).
  - flow_rate: "the pores are typically so large that the flow is only constrained by the grind size" (Gagné, verbatim).
  - compostability: a reusable insert is not thrown away per brew; its value is `not compostable` by the register's definition. Chain: reusable (filter_material definition) → not disposed of per brew → compost route not part of the choice.
- when `cloth`: then_ask cholesterol_oils_filtered (shown, not asked); makes_irrelevant bleaching, whitening_method, perforated, flow_rate, compostability. (compostability added 2026-10-25.)
  - cholesterol_oils_filtered: cloth sits between paper and mesh (Orrje fabric filter; Naidoo sock; Wikipedia "allows more of the oil to come through than paper filters").
  - flow_rate: "Cloth filters suffer from a similar problem to the metal filters in terms of flow rate and large pores" (Gagné, verbatim).
  - compostability: definitional. The register defines it as "Whether a paper filter is certified ...", so it cannot apply to cloth. The earlier note (cloth wears out and is plant fibre) was not sourced and contradicted the definition.
- when `paper`: then_ask bleaching, compostability.
  - bleaching: split 3 follows only for paper by definition.
  - compostability: Melitta's "Commercially Compostable. Facilities may not exist in your area" and BPI's two schemes (verbatim).

**Split 3, bleaching ("White filter paper, or brown that needs a good rinse?")**

- when `white`: then_ask whitening_method (shown, not asked; the facet is not novice-answerable).
  - Melitta's cones say "Chlorine Free**" with "** Elemental Chlorine Free" while Chemex ("oxygen cleansing bath") and Hario Europe ("oxygen bleached") say oxygen for white (verbatim); "chlorine dioxide ... is used alone in ECF" (Wikipedia, verbatim). So "white" covers two processes the buyer may care about.
- `brown (unbleached)` → makes_irrelevant whitening_method: removed 2026-10-25. Melitta's "#4 Cone Filter Paper Natural Brown" page carries the same "** Elemental Chlorine Free" footnote (products.json, verbatim) and If You Care labels its brown paper "totally chlorine-free (TCF)" (verbatim), so brown listings do carry the labels and the facet is not irrelevant for them. Chemex's "The brown filter paper goes through only the hot water baths" still holds for Chemex alone.

## Dropped

- Paper thickness as its own facet. It is the advertised number, but two sources show it does not predict flow; it lives in the flow_rate judgement.
- Grammage (g/m2). Rendón measured 50 to 67 g/m2, Wikipedia says about 100; nobody lists it.
- Tabbed versus tabless V60 papers. Gagné found the tabless Hario thicker, slower and more clog-prone, but it is one brand's two SKUs.
- Mesh count for metal filters ("600 Mesh" in some Amazon titles). Present on listings, but I found no expert source on what mesh count does to the cup, so it is not a facet.
- Fibre (abaca, bamboo, wood). Cafec and Hario Meteor sell abaca for flow, Melitta sells 40 percent bamboo for sustainability; no measured effect beyond what flow_rate already holds.
- Burst resistance (double crimp, "never break" seam). Advertised, never tested in what I read.
- Pack count as a facet. It fails the generic test; it is used only as the listing field behind cost per filter.
- The 1980s dioxin history. Real, but I could not fetch a primary source after the search budget ran out, and the modern ECF/TCF definitions carry the buyer-relevant part.

## Trimmed detail

- `flow_rate`: Gagné's full order of V60-fit papers, slowest to fastest: Cafec 5.79, Hario tabless 6.89, Chemex bleached 7.23, Chemex unbleached 9.82, Hario tabbed 11.03, Hario unbleached 15.3, V60 cloth 18.1, Osaka metal 67.8 mL/s. Unbleached was faster than bleached within both Hario and Chemex. Hario Meteor (not in Gagné's 2019 test) claims "nearly 3 times the permeation rate of the classic V60 filters".
- `bleaching`: Gagné's 150 g hot-water rinse cleared an unbleached Hario 02 completely; tap-water rinsing left small amounts. Bleached papers gave 0 ppm unrinsed.
- `cholesterol_oils_filtered`: Orrje's medians put paper-filtered home coffee at about an eighth of French press and percolator (12 versus about 90, same units) and fabric-filtered boiled coffee at 28 against 939 unfiltered. Naidoo's Indian metal-mesh coffee (0.05 mg/cup) was as low as paper, while Indonesian metal-mesh coffee (0.98 mg/cup) matched espresso.
- `perforated`: Rendón compared filters from Brazil, Japan, the US, Germany, France and the Netherlands with one high-cafestol Catuai coffee at 1:10; perforated papers gave the highest diterpene levels among similar-sized filters.
- `filter_size`: Wikipedia's European 1× series (1×2, 1×4, 1×6) shares one wedge geometry at different heights (bottom width about 49 mm, angle about 54°, one source). Only US #4 and #6 match 1×4 and 1×6; US #2 is "similar to 102", a different, larger shape than 1×2, so the register names only #4 and #6 as equivalents.
- `compostability`: Hario Meteor's PLA layer; BPI's home scheme follows NF T 51-800 and commercial follows ASTM D6400.

## Corrections after the blind check (2026-10-25)

Verified against `check.md`'s 18 required fixes. Most had already been applied to `category.json` and `notes.md` (the Branches trace above carries inline "2026-10-25" change markers for each); this pass re-verified each one against the live files and made the one remaining correction (item 1).

1. Food-contact judgement: the `judgements` entry and its note already state BfR XXXVI/1 is a non-binding German recommendation, not an EU legal norm. Fixed now: the `bfr_paper_2018` reference's `supports` line still read "covers hot filter papers ... under EU 1935/2004," repeating the misframing; changed to "BfR XXXVI/1 is a non-binding German recommendation, not an EU legal norm." 21 CFR 176.170 left as is.
2. bleaching split rationale: already reads "Over a quarter of Amazon paper titles omit colour, so bullets or the photo fill it," not "every listing." Confirmed the "appear in nearly every title" line in the Amazon source note above is already corrected to "colour and count are often missing."
3. White papers and rinsing: summary, bleaching `why_experts_care`, and the rinsing judgement already say Gagné measured dissolved solids, not taste, on one brown paper, and still advises pre-rinsing white paper. No further edit needed.
4. "Unbleached papers drained more evenly": tradeoffs entry already reworded to "had more uniform pores in one tester's microscope analysis."
5. Brown → makes_irrelevant whitening_method: leg already absent from the bleaching split's branches; `whitening_method`'s `why_experts_care` already says Melitta footnotes white and brown cones alike.
6. 1×4/1×6 equivalence: already removed from values, vocabulary and the numbers table (single-source, Wikipedia hedges with "similar to").
7. Citation URLs: `chemex_faq`, `hario_usa_flow`, `cafec_th3`, and `melitta_basket` already added as separate references; `hario_usa_v60`'s title already scoped to "02."
8. Wave 155/185: upgraded past "mark as single-source" — `kalitausa.com` (not the dead `kalita-usa.com`) answered on 2026-10-25, so both sizes now carry a Kalita USA source (`kalita_usa_wave155`, `kalita_usa_wave`) alongside Hario Flow and Amazon.
9. AeroPress leg: already dropped from `brewer_fit` branches; `flow_rate`'s `why_experts_care` already says pressing harder makes flow "matter less," not irrelevant.
10. Non-askable then_ask targets: `flow_rate`, `cholesterol_oils_filtered`, and `whitening_method` already end their `why_experts_care` with "Shown, not asked." The V60 leg's flow_rate is consistent with the facet's own backfill note, which covers V60 and Chemex only (wave is not asked).
11. Cloth and compostability: `compostability` already added to both mesh and cloth branches' `makes_irrelevant`.
12. filter_material backfill: already "low," not "none"; the listing hint already flags charcoal water filters as an out-of-scope false match, and the backfill note already covers "Reusable"/"Goldtone" titles omitting material.
13. Scope boundary: a new judgement already states the scope (standalone drippers to `manual-coffee-tea-prep`; refillable pod cups and charcoal filters out) in `category.json` itself, not only notes.md; `brewer_fit`'s value already reads "paper pod insert or drip bag" so it cannot be read as a refillable K-cup.
14. filter_size values: already replaced the single-sourced "4 to 6 cup basket" with the generic "smaller basket (cup range as stated)," covering the 1-4/4-5/2-5 cup listings without inventing a number; percolator disc/wrap and pod insert values already present.
15. cost_per_filter: `why` already notes that titles such as Melitta's "N-Pack" omit per-pack count and that it has to come from the bullets; the recount above is already corrected.
16. whitening_method `why_experts_care`: already states researchers disagree on whether TCF effluent is less harmful than ECF, sourced to `wiki_pulp_bleaching`.
17. Hedges: `cholesterol_oils_filtered` already qualifies gold-tone mesh with "though one study found some metal-mesh coffee as low as paper, so values are typical for the material, not measured" (the "presumably" hedge itself is carried in the `nutritionfacts_filter` reference's supports line); `perforated` already adds "the spread across all its papers was narrow, so the difference is small."
18. bleaching question wording: already carries the coffee-specific hook "or brown that needs a good rinse?" rather than a bare colour question.

No findings rejected; all 18 are satisfied in the current `category.json` and `notes.md`.

## Script and schema problems

- `schema.json` restricts `vertical` to formulations, machines, fashion, furnishing or other. The brief asks for `vertical: supplies`, which the validator rejects as a schema error. I set `vertical` to `other` so validate, render and viewer run clean, and report the conflict; the catalog entry keeps `vertical: supplies`. A later step can switch the value once the schema accepts it.

## Corrections (2026-10-25, author's STYLE pass before hand-off)

1. `vertical` set to `other` (validator rejects `supplies`; see "Script and schema problems").
2. Summary trimmed from 61 to 58 words; the white-versus-brown sentence shortened.
3. `filter_material`: removed "Gagné measured a metal cone draining several times faster than any paper" (a single-source ratio); the claim now says the holes are open enough that grind alone sets brew time (Gagné verbatim).
4. `cost_per_filter`: removed the "40 to 1,000" pack range (two endpoints, one source each).
5. `flow_rate` and the flow trade-off: removed "fast paper suits finer grinds" and "a fast paper is forgiving", which no source says; replaced with Chemex's too-fine-grind clogging line and Hario's clarity claim for Meteor.
6. `perforated` definition no longer says the holes are "to speed flow"; Melitta says they let "the full coffee flavor" through.
7. `filter_size` and the Wikipedia reference: corrected "1×2, 1×4 and 1×6" to "#4 and #6 match 1×4 and 1×6"; Wikipedia maps US #2 to 102, not 1×2.
8. `cholesterol_oils_filtered`: replaced the unsupported "outranks taste" opinion with who the facet matters to.
9. Visual grammar "Clarity to body scale" has no `origin` (it is a proposal); the Coffee ad Astra map names its origin without a trailing period.
10. Mechanism: glossed "creping" where it first appears.