# Cotton Balls, Pads & Swabs: research notes Date: 2026-10-25. Scope: disposable (and a minority of reusable) absorbent-cotton personal-care consumables shaped as loose balls, flat rounds/pads, or fiber wrapped on a stick (swabs/buds/cotton-tip applicators), used for skincare and makeup removal, general hygiene, baby care, and minor first aid (cleaning a scrape, applying antiseptic, packing a nosebleed). Not in scope, and left to siblings in `catalog.json` under `supplies > personal`: feminine hygiene pads (`feminine-pads`), gauze, adhesive dressings and medical tape (`medical-tapes`), assembled first-aid kits (`first-aid`), and makeup-application sponges/brushes (`makeup-brushes`). Craft-grade polyester fiberfill balls sold for decor or stuffing are also out of scope; they are a different fiber and a different vertical (crafts), not personal care. Quote labels: **verbatim** = I saw the exact sentence in raw text I fetched myself (an API's JSON/XML field, or HTML I grepped after stripping tags). **fetched excerpt** = the fetch tool returned a model's summary/paraphrase of a page it actually loaded. **search excerpt** = from a search-engine summary, page itself unseen. No search-excerpt-only numbers are in category.json. WebSearch reported its session budget exhausted on the first three queries of this run, before any query returned results, so this entire register was built from direct fetches (WebFetch, curl), a REST API (Europe PMC, which returns raw abstract/full-text JSON and XML, not a search summary), and one Internet Archive availability lookup. Retail listings (Target, CVS, Amazon-style) and several brand/manufacturer sites (Dukal, McKesson, Medline, Grainger, Puritan, Q-tips/Unilever, Mayo Clinic, EUR-Lex, FDA, CDC, entnet.org, healthychildren.org, Allure, Byrdie, Paula's Choice, two lash-supply sites) all returned 403/404 or could not be reached; I could not confirm any field is present on a retail listing in this run, so nothing is marked `source: listing` on that basis alone. ## Seed Seed run (recorded verbatim from the script output): ``` === Example entry (from an unrelated category) === category: bedsheets what this shows: 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. [... weave facet JSON ...] === Persona === A pastry chef, who thinks in ratios, temperatures and why a recipe fails ``` 1. What is the cotton-balls equivalent of the example? The generic number everyone advertises is "100% cotton" on the pack. The real split underneath it is construction, not purity: whether the fiber is left loose (soft, absorbent, sheds lint) or carded and bonded into a pad or a lint-free blend (holds together, sheds less), and separately whether the batch is sterile and individually wrapped or bulk and non-sterile. Two "100% cotton" products can behave completely differently depending on these, the same way two sheets at the same thread count feel different by weave. 2. What would the pastry chef notice first? A pastry chef thinks about process, not ingredient ratio alone: identical flour and sugar percentages still bake differently depending on how the batter was mixed. Applied here, they would ask how the cotton was processed and packaged, not what percentage of it is cotton, which is exactly the fiber-construction and sterility point above. Kept: fiber construction (loose vs. bonded/lint-free) and sterile-vs-bulk as the real axes, instead of a bare cotton-percentage claim. Dropped: no facet called "cotton purity %" — no source gives this as a real spec, and it reads as a label claim, not a behaviour difference, unless construction is also known. ## Sources and what each supports 1. **Jatana et al., "Pediatric Cotton-Tip Applicator-Related Ear Injury Treated in United States Emergency Departments, 1990-2010"** (Journal of Pediatrics; PMID 28473166), read via the Europe PMC REST API (`resultType=core`), which returns the publisher's own abstract text as a JSON field — **verbatim**. Supports: 263,338 US children under 18 treated for cotton-tip-applicator ear injuries, 1990-2010; highest injury rate age 0-3 (32.2 per 100,000); ear cleaning was the documented circumstance in 73.2% of cases; the patient was handling the swab in 76.9%. Kind: academic. 2. **"Unpacking cerumen impaction: a systematic review of clinical practice guidelines... WHO package of ear and hearing care interventions"** (PMC 13130458, 2025), read via Europe PMC's `fullTextXML` endpoint, raw XML text — **verbatim**. Supports: "over 90% of individuals engage in ear cleaning practices... such as inserting foreign objects (i.e., cotton-tipped swabs)"; a guideline statement, "Advise adults not to remove earwax to clean their ears by inserting small objects, such as cotton buds, into the ear canal, as this could damage the eardrum and canal and push wax further down into the ear." Kind: academic. 3. **AAO-HNS, "Clinical Practice Guideline (Update): Earwax (Cerumen Impaction)"** (PMID 28045591), abstract via Europe PMC — **verbatim** abstract text, but the abstract itself (as returned) names ear hygiene education as a recommendation without naming cotton swabs specifically; I did not get the full guideline body, which the 2025 systematic review above quotes as containing the "advise... not to insert... cotton buds" line. Used only to support that a standards body (AAO-HNS) has an active cerumen guideline that the 2025 review is summarising; the specific wording is attributed to source 2, not read directly in source 3's own text. Kind: academic. 4. **European Commission, "Single-use plastics" topic page** (environment.ec.europa.eu), fetched live — **fetched excerpt** (WebFetch summary of the live page). Supports: "Cotton bud sticks" is listed as the first of the "10 single-use plastic items" targeted by the EU Single-Use Plastics Directive; "3 July 2021 - EU no longer allows certain SUP items to be placed on Member States' markets." The page's own text, as returned, did not explicitly re-confirm that cotton bud sticks are among the items that specific date applies to (it names the date for "certain SUP items" generally); I treat the date as applying to cotton bud sticks on the strength of source 5 below, not from this page's wording alone. Kind: regulator. 5. **Wikipedia, "Cotton swab"** (redirects from "Q-tips"), fetched live twice (same article both times) — **fetched excerpt** both times, consistent. Supports: medical swabs run longer and single-tipped, packaged sterile one or two to a sleeve, because paper/wood sticks survive autoclaving and plastic would melt; home swabs run shorter, double-tipped, sold non-sterile in packs of 100+; several countries (Italy 2019, Monaco 2020, England/ Scotland/Wales/Isle of Man 2019-2021) restricted plastic-stem swabs ahead of or alongside the EU-wide rule. The specific centimetre/inch figures and the country-by-country dates are **single-sourced to this one encyclopedia article** (fetched twice is not two sources) and are therefore described in category.json only in relative, non-numeric terms ("longer", "shorter"), never as a sourced range. Kind: primer. 6. **Wikipedia, "Cotton pad"**, fetched live — **fetched excerpt**. Supports: cotton pads are used medically to control bleeding from punctures, injections and blood draws, and for infant hygiene, alongside cosmetic makeup use; a 1965 industry concern that "synthetic puffs" were being sold labelled as cotton balls. Used only for the qualitative uses and as evidence that fibre content has been a real point of concern in this market before; the market-size dollar figures in this article are not used anywhere in category.json (single-sourced, and not load-bearing for any facet). Kind: primer. 7. **Wikipedia, "Nonwoven fabric"**, fetched live — **fetched excerpt**. Supports only the general manufacturing fact that "rayon used to be a common fibre in nonwovens, now greatly replaced by polyethylene terephthalate (PET) and polypropylene," i.e., that bonded/blended cotton-alternative nonwovens are a real, named manufacturing category. Did not yield a lint or wet- strength comparison; I dropped that line of claim rather than assert it unsourced. Kind: primer. 8. **Swisspers**, brand site, fetched live — **fetched excerpt**. Supports: the brand sells a separate "Ultra-Luxe Cotton Round" line marketed as "100% cotton, lint-free," distinct from, and more expensive than, its standard round; and a "Premium Exfoliating Cotton Round" that is dual-sided, one smooth face and one textured face for light exfoliation. This is the only source for both the lint-free-as-a-separate-line claim and the dual-texture claim; both are flagged medium/low confidence in judgements and the `fiber_material` and `pad_texture` facets below for that reason. Kind: spec_sheet. 9. **LastObject, "LastSwab"**, brand site, fetched live — **fetched excerpt**. Supports: a reusable, washable swab product exists and is marketed specifically as a replacement for single-use cotton swabs. The brand's own multiplier and cost-per-use figures ("1,000+ cotton swabs," "8.3x less environmental impact," "$0.01 per use") are single-sourced marketing claims from the seller and are not used anywhere in category.json. Kind: spec_sheet. 10. **EcoRoots, reusable cotton rounds**, brand site, fetched live — **fetched excerpt**. Supports: a washable organic-cotton-and-bamboo round exists and is marketed as a replacement for disposable cotton pads, is not pre-washed (shrinks on first wash), and is reusable over "a couple of years" per one quoted customer review (anecdotal, not a controlled wash-count, so no wash count appears in category.json). Kind: spec_sheet. Sources tried and not reachable, listed because the brief asks what source types I could not reach: ASTM (D4391 catalogue page, 403), the FDA consumer page on cotton-tip applicators (the exact URL could not be found; FDA's own site returned 404 for every guessed slug and has no public search I could fetch), CDC, Mayo Clinic, healthychildren.org, entnet.org (all 404/403), EUR-Lex itself (returns HTTP 202 with an empty body to this tool, twice, both direct and via archive.org's availability check, which found no snapshot), Dukal, McKesson, Medline, Grainger, Puritan Medical Products (medical-supply spec sheets, all 403/blocked), Q-tips/Unilever's own brand page (403), Target and CVS retail listings (404/403), Allure, Byrdie, Paula's Choice, NAILS Magazine (no usable content returned), and two lash-supply sites that failed DNS resolution entirely. USP's own "Purified Cotton" / "Absorbent Cotton" monograph was searched for via Europe PMC and not found as a fetchable, citable text; I did not put any USP-specific number in category.json as a result, and the sterile-vs-bulk distinction is sourced to source 5 (Wikipedia, for swabs) and extended to balls and pads by stated analogy in judgement 2 below, not independently confirmed for balls/pads. ## Data points used for derived metrics None. I considered a price-per-unit or cost-per-use metric, but it is the same arithmetic (price divided by count) a shopper would do for any multi-pack commodity, and reads unchanged for paper towels, battery packs or diapers — it fails the generic test and is not expertise, so `derived_metrics` is omitted rather than padded with a non-metric. ## Against the existing add-on schema Add-on read: `supplies/personal/add-ons/cotton_ball.jsonc`, which is empty (`{}` under a comment naming the category) — there is no field to compare against. Composed with `personal-standard.jsonc` (`replacement_compatibility_raw`, `sizing.weight_capacity`, both inapplicable: nothing here is "replaced" and nothing bears weight) and `supplies-standard.jsonc` (`high_signal_markdown`, `sizing.size_stated`, `sizing.pack_count`, `sizing.dimensions`, `sizing.capacity`, `sizing.weight`), which already cover pack count, size label and dimensions generically, so none of this register's facets duplicate them. Applying AGENTS.md rule 1 (visually-derived fields stay out of the schema) turns out to cut most of the candidate facets below from the add-on, because this is a visually simple product: a ball, a flat pad, a dual-texture pad and a double- vs single-tipped swab are all things a photo answers as well as text does, so they belong to the image pipeline, not a new schema field, even though they are real facets in the register below (the register is about what the assistant should ask or show, not only what the text pipeline extracts). | Facet in this doc | In add-on today | Proposed kind under AGENTS.md | Note | |---|---|---|---| | format (ball / pad / swab) | no (add-on is empty) | none — cut, photo-derivable | A listing photo shows a ball, a flat pad or a stick as reliably as any bullet does; the taxonomy leaf itself already separates the product type at the catalog level. Not a new field. | | sterile_or_medical_grade | no | hard filter (boolean) | Not visible in a photo reliably; "sterile," "individually wrapped" is a text claim on the package and the listing bullet. Worth a real field. Researched, low backfill cost (brand/medical-supply pages state it plainly; I could not confirm it is reliably in retail-listing titles in this run). | | fiber_material (100% cotton vs. lint-free blend) | no | high_signal_markdown, not a filterable enum yet | Only one brand (Swisspers) was confirmed to market this as a named, separate line; too thin an evidence base to commit to as a dense, backfill-worthy enum today. Belongs in the markdown line until more brand pages confirm the pattern. | | stick_material (paper/wood vs. plastic, swabs only) | no | hard filter (enum), swab rows only | Null for ball/pad rows. A real buyer filter in the EU (plastic-stick swabs cannot be newly sold there since 3 July 2021) and for anyone avoiding plastic. Researched; brand/manufacturer pages state it, not confirmed on retail listings here. | | tip_count (single vs. double, swabs only) | no | none — cut, photo-derivable | Visibly countable from the product photo. | | pad_texture (smooth vs. dual-sided textured, pads only) | no | none — cut, photo-derivable | The waffle/ridge texture is visible in the product photo; also single-sourced (Swisspers only) in this run, so even as a markdown line it should wait for a second source. | | reusable_or_disposable | no | none — cut, photo-derivable | A terrycloth round or a silicone swab looks obviously different from a disposable one in the product photo; the taxonomy/title ("reusable," "washable") also carries it, so no new structured field earns its keep here. | Nothing in today's add-on is called out as noise, because today's add-on holds nothing. The risk this register flags for whoever fills `cotton_ball.jsonc` next is the opposite one: most of the obvious-looking facets for this category are exactly the kind AGENTS.md rule 1 says to leave to the image pipeline, so the temptation to add format/texture/tip-count columns should be resisted; the two fields actually worth adding are `sterile_or_medical_grade` and `stick_material`. ## Branches Each leg names the sentence in the register or these notes it follows from. The `when` values on `format` are its three facet values, worded to match the split question's three answers 1:1 ("cotton balls" -> `cotton ball`, "flat pads" -> `round or square pad`, "swabs on a stick" -> `swab on a stick`). The `when` values on `sterile_or_medical_grade` are `yes` / `no`, per the schema's boolean convention. **Split 1, format ("Cotton balls, flat pads, or swabs on a stick?")** - when `cotton ball`: makes_irrelevant `stick_material`, `tip_count`, `pad_texture`. "A loose ball has no stick and no second flat face" (format, definition) — there is nothing for those three facets to describe on this branch. - when `round or square pad`: then_ask `pad_texture`, `reusable_or_disposable`; makes_irrelevant `stick_material`, `tip_count`. "A pad's two faces can be smooth or one can be textured" (pad_texture, definition), so the question applies here and nowhere else; "a washable terrycloth round... replaces disposables" (reusable_or_disposable, why_experts_care) names pads by name. Sticks and tip count do not exist on a flat pad. - when `swab on a stick`: then_ask `stick_material`, `tip_count`, `reusable_or_disposable`; makes_irrelevant `pad_texture`. The stick and tip facets are definitionally about the stick format; "a reusable, washable swab... replace[s] single-use cotton swabs" (reusable_or_disposable why_experts_care, sourced to LastSwab) also names swabs by name. A stick format has no second textured face, so `pad_texture` does not apply. **Split 2, sterile_or_medical_grade ("For a wound or baby's skin, or just toiletries?")** - when `yes`: makes_irrelevant `reusable_or_disposable`, `fiber_material`. Chain: sterility is what the "for a wound or baby's skin" answer is buying (facet definition); "it must be laundered between uses... is never sold sterile" is stated of the reusable facet's `why_experts_care`, so a reusable item cannot satisfy the sterile answer and the question stops being useful once sterile is required. Separately, `fiber_material`'s own `why_experts_care` says lint only matters for cosmetic precision tasks ("wet nail polish, lash glue and camera lenses"), none of which applies once the buyer needs sterile wound/baby-skin cotton, so lint stops being a reason to ask or filter on that branch too (added 2026-10-25, see Corrections below). - when `no`: then_ask `fiber_material`, `reusable_or_disposable`. Nothing about everyday toiletry use forces either answer, so both remain open questions — this leg exists to say they are still worth asking, not to narrow anything, matching the house convention (e.g. power-bank's "no branch for `none`" leg) of recording a no-op leg rather than omitting it silently. No branch on `fiber_material`, `stick_material`, `tip_count`, `pad_texture` or `reusable_or_disposable` downstream, because none of them is a split and nothing in the register states that any of their values changes what else gets asked. ## Trimmed detail - The Wikipedia "Cotton pad" US market-size figures ($343.1M in 1999 to $177.7M in 2014; brands Swisspers, Swiss Beauty, Cotton Cloud in 2016) and the 1937 Joseph A. Voss pad-cutting machine are real but single-sourced and not load-bearing for any facet or judgement; kept only here for colour. - LastSwab's own multiplier and cost claims ("1,000+ cotton swabs," "8.3x less environmental impact," ISO-certified LCA, "$0.01 per use") are a single seller's marketing figures. `reusable_or_disposable`'s `why_experts_care` says only that it "replace[s] hundreds of disposables," not the seller's specific multiple. - The precise swab lengths from Wikipedia (medical ~10 cm/4 in single-tipped; home ~6.5 cm/2.5 in double-tipped) are single-sourced to one article fetched twice. `tip_count`'s facet text keeps the single/double distinction (which is also what actually matters for the child-safety story) and drops the centimetre figures. - Country-by-country plastic-swab ban dates (Italy 2019, Monaco 2020, England/ Scotland/Wales/Isle of Man 2019-2021) are from the same single Wikipedia source; `stick_material` cites only the EU-wide, two-source date (3 July 2021). ## Corrections First pass; no blind check has run against this register yet. Flagging two things for a checker to look at first, since I could not fully resolve them in this run: (1) `fiber_material` and `pad_texture` both rest on one brand's site (Swisspers) and should be checked against a second cosmetic-cotton brand before being trusted at more than low/medium confidence; (2) the `sterile_or_medical_grade` facet's claim that bulk toiletry cotton balls/pads are generally non-sterile is extended by analogy from the swab-specific Wikipedia sentence, not independently confirmed for balls and pads, and should be checked against a medical-supply brand page (Dukal/McKesson/Medline all blocked this run) if one becomes fetchable. ## Corrections after the blind check (2026-10-25) 1. Added `fiber_material` to `sterile_or_medical_grade`'s `yes` branch `makes_irrelevant` list in `category.json`, alongside `reusable_or_disposable`. Fixes required-fix #1: the facet's own `why_experts_care` says lint only matters for cosmetic precision tasks (nail polish, lash glue, camera lenses), none of which applies once sterile wound/baby-skin cotton is needed, so a sterile-branch flow had no stated reason to still ask or filter on fiber material. Updated the matching leg in this file's "Branches" trace for Split 2, `when: yes`. 2. Added a sentence to `tip_count`'s and `pad_texture`'s `backfill_note` in `category.json` naming the product photo as a cheaper backfill route than brand-page text research. Fixes required-fix #2: `category.json` alone (the file the pipeline consumes) previously sent a reader hunting brand pages when the photo route, already worked out in this file's "Against the existing add-on schema" table, is cheaper and reliable. Did not add a fourth `source` enum value ("derivable from image") because `schema.json` is outside `cotton-balls/` and out of scope for this fix; flagging here that the schema change the checker suggested as a possible next step is still open. 3. Added the clause "the directive's general application date" to `stick_material`'s `why_experts_care` next to the "3 July 2021" date. Fixes required-fix #3: the fetched EC page never prints "cotton bud sticks" and "3 July 2021" in the same sentence; the day comes from the Single-Use Plastics Directive's general application date (confirmed by source 4 in this file), not from text naming the item on that date. The hedge keeps the date (true and two-source-backed at the year level, per Part A claim #2 of `check.md`) while flagging its exact-day provenance so a future checker does not have to re-derive the chain. 4. No change made for required-fix #4 (the 1965 "nylon puffs labelled as cotton balls" single-source line). The checker marked this informational only: it is not cited by any facet, split, branch or judgement, so it carries no risk today. Rejected as a fix target for that reason; left in `category.json`'s `mechanism` evidence and `wikipedia_cotton_pad`'s `supports` line unchanged, with this note as the flag against future promotion to a load-bearing claim without a second source. No new web research was needed for any of the four fixes; all were textual corrections to claims and backfill notes already sourced elsewhere in this register. ## Build notes `validate.py`, `render.py` and `viewer.py` ran clean on the first attempt with no script misbehaviour to report.