Fact-check / Corrections
Corrections and re-checks
8 ratings changed and 8 re-checks that held, across 83 published verdicts. Nothing here replaces what was first published; each verdict page keeps the original reasoning under its re-check trail.
Think a verdict is wrong? Write to hello@nozzlecheck.com with the claim and what the source says. How ratings are decided is on the method page.
Ratings changed
- ◐ Partly true → ? Figure unverified“A Delta E 2000 value of 1 is the threshold below which a colour difference is imperceptible.”
We rated this Partly true and called the 2024 study the better authority. An outside audit pointed out that the study measured something else. It tested ceramic dental shade tabs, viewed by 20 dentists and 20 patients, and the threshold it found is the point where half of observers see a difference. That is a different material and a different viewing task from print, so it cannot set or contradict a threshold for printed colour. The '1' figure comes from a vendor blog with no data or method, and our sources hold no measurement made on printed samples. We confirmed that the vendor says 1, not that 1 is the threshold, so the figure is unverified.
pmc.ncbi.nlm.nih.gov → - ✗ False → ◐ Partly true“TPU hot-melt DTF adhesive powder is marketed as "non-toxic, odorless, eco-friendly, and non-irritating to human skin" and "harmless to the human body and the environment."”
We rated the marketing false on one other seller's warning, for a different product, and said that warning was backed by a safety data sheet. The DTF Station page carries the warning but links no SDS; its specifications give polyurethane as the main raw material. Two SDSs for TPU DTF powders, read on 2026-10-24, find no hazard classification for their own products while still calling for dust or inhalation precautions. (Wording corrected 2026-10-25: this note first said the DTF Station page does not name the polymer, and that both SDSs say to avoid dust; only Wellington House does.) The absolutes in the promotion are still unsupported; 'false' for the whole claim was not.
d2zn16t8uygl6t.cloudfront.net → - ◐ Partly true → ✗ False“Kornit's NeoPigment inks, as used by Printful, are non-toxic and adhere to OEKO-TEX STANDARD 100 and the Global Organic Textile Standard (GOTS).”
Kornit's own safety data sheet for NeoPigment Eco-Rapid White ink (SDS 60-SDSU-0112, revision 4.2, 9 April 2024) classifies it Acute Tox. 4, H302 "Harmful if swallowed", and Skin Sens. 1, H317 "May cause an allergic skin reaction", with the signal word warning, under 29 CFR 1910.1200. The ink's maker contradicts "non-toxic". The two certification marks were already cited at the wrong object when this verdict was published: STANDARD 100 certifies the garment and GOTS approves inputs without declaring them free of hazardous substances. No part of the claim holds as worded, so the rating is False. The SDS covers the white ink, the one a Kornit DTG shop uses most; we have not read the sheets for the other colours. kornit.com refuses our fetcher, so the sheet was read with a browser user agent.
kornit.com → - ◐ Partly true → ✗ False“Hardware breakdowns on Epson SureColor F2100 printers requiring damper assembly replacements and localized repair support”
We rated this Partly true while our own reasoning refuted every part of it. The thread is one owner of one used F2100 with clogged white ink lines and a clogged printhead, whose damper-assembly problem began when they removed two screws during their own repair. That is not a pattern of breakdowns. No damper was replaced or ordered. The thread now has one reply, on 30 August 2026, from the forum's AI community assistant asking for photos, which is not repair support of any kind, local or otherwise. We also said the thread had 0 replies; it has that one. The source the claim was taken from contradicts it: False.
t-shirtforums.com → - ? Figure unverified → ✓ Confirmed“Under the US Copyright Office's 2023 registration guidance, material generated entirely by an AI system from a text prompt is not the product of human authorship and cannot be registered, while a human's own selection or arrangement of that material can be.”
Read in full, the 16 March 2023 statement of policy (88 FR 16190) settles both halves; the first check read a copy cut off in Section II. Where an AI technology "determines the expressive elements of its output, the generated material is not the product of human authorship" and must be disclaimed in an application. The same section then says a human may select or arrange AI-generated material in a sufficiently creative way that the work as a whole is original, and that copyright then protects only the human-authored aspects. The Copyright Office's own copy at copyright.gov/ai/ai_policy_guidance.pdf carries the same text. One condition the claim leaves out: the selection or arrangement has to be sufficiently creative, and the AI-generated material inside it stays unprotected.
federalregister.gov → - ✗ False → ◐ Partly true“A seller who uploads a prompt-only AI image to Printful can truthfully warrant, as Printful's print file content policy requires, that they own the rights to it.”
We rated this False on the wrong reading. Printful asks an uploader to confirm the rights to use, sell and distribute a print file. It does not ask them to warrant that they own a copyright in it. The Copyright Office's 2023 statement says material whose expressive elements an AI determined is "not protected by copyright", which rules out claiming to own a copyright in a prompt-only image. Nothing in it stops anyone using, selling or distributing material that nobody owns. So the warranty Printful actually asks for is one a seller of a prompt-only image can give, and the claim is right about that. It is wrong about what the warranty says: "that they own the rights to it" is not Printful's wording. Verdict 78, which reads the same Printful sentence this way, stands. Neither source covers the AI tool's own terms of use, which decide whether its output may be sold, or output that copies someone else's protected work. Printful has since moved the article to a new address; the wording is unchanged.
help.printful.com → - ◐ Partly true → ✓ Confirmed“Enfocus says Griffin's true-shape nesting and gang-run printing cut material costs by up to 50%.”
The first check said the Griffin page gives up to 50% only for waste and lists cutting costs as a separate benefit without a percentage. The live page states the claim as worded: under 'Why print professionals choose Griffin' it says True Shape Nesting and gang-run printing cut material costs by up to 50%. Enfocus's own figure, with no stated method; the rating confirms what Enfocus says, not that the saving is typical.
enfocus.com → - ✓ Confirmed → ? Figure unverified“Curing DTG prints takes 1.8 to 2.4 kWh per 100 garments and adds 11 to 19 cents a shirt at $0.12 per kWh, according to packprintmatrix.com.”
The two figures in the claim cannot both be true. At $0.12 per kWh, 1.8 to 2.4 kWh per 100 garments costs $0.0022 to $0.0029 a shirt, not $0.11 to $0.19. The second figure would need 0.92 to 1.58 kWh per shirt, 38 to 88 times the first. The page states both and gives no method for either, and no second source has been found. We rated the claim Confirmed because the page says it; that confirmed the attribution, not the figures, and neither figure is verified. This verdict ran as Verdict of the week in Pulse 2026-10-06.
packprintmatrix.com →
Re-checked, rating held
- held ✗ False“FDA's ceramic-mug lead rule (CPG Sec. 545.450) mandates keeping decoration off a 20mm zone at the rim/lip of a mug to avoid oral ingestion of decorative components.”
An outside audit pointed out that this verdict rated the claim False while saying no text of FDA CPG Sec. 545.450 was in our sources. Under our own definitions that is No evidence, not False. We have now read the CPG in full. It sets limits on lead leaching from food-contact surfaces, measured by ASTM C738 on six units: 0.5 micrograms per mL for any one cup or mug. It defines flatware and hollowware by their depth below the rim, and it says nothing about decoration near the rim or a 20 mm zone. The rule the claim names requires a leach limit, not a decoration-free rim, so the rating stays False, now resting on the FDA text. The 20 mm figure comes from Canada's SOR/2016-175, which allows rim decoration within limits, as the original reasoning says.
fda.gov → - held ✓ Confirmed“EmpCo transposition across the EU is incomplete as its 27 September 2026 application date approaches: 'several countries have not yet completed transposition, the legal position is still developing across...”
Re-read on 2026-10-26: the CMS tracker, now marked updated on 21 September 2026, still carries the sentence. The 27 September 2026 application date has since passed; the claim and the first check describe the position before it.
cms.law → - held ✗ False“DTG files should be CMYK to prevent unexpected colour shifts on the finished product”
The rating stands; our quote did not. Fourthwall reworded its colour-profile line after we published: the clause "converting it yourself can shift colors" is no longer on the page. The quote and the reasoning now carry its current wording, which still says sRGB and no conversion to CMYK. Printful's page, re-read the same day, still says converting to CMYK before uploading can reduce the colour information.
help.fourthwall.com → - held ◐ Partly true“The standard DTG ink curing temperature is 330°F (165°C) for 90 seconds.”
The rating stands; one row of our reasoning did not. We gave Image Armor 350°F for 35 seconds, which is Winnerjet's table, not the ink maker. Image Armor's own page gives its E-Series white ink a 35-second cure at 356F with light to medium pressure. The spread we described is wider than we said, not narrower.
imagearmorpt.com → - held ◐ Partly true“The standard DTG ink curing temperature is 330°F (165°C) for 90 seconds.”
Our DuPont row was Winnerjet's too, and it does not match DuPont. DuPont's own process guidelines for Artistri Brite P5000 ink cure dark or black cotton on a heat press for 60 to 75 seconds at 356°F (180°C), or 35 seconds at 365°F with a non-stick sheet. The 330 to 345°F range appears only for light shirts with a light white underbase, at 60 to 90 seconds. DuPont's figure for the shirts that carry white ink is 26°F hotter than the "standard" this claim names, which is more evidence for Partly true, not less.
dupont.com → - held ✓ Confirmed“In a poly/cotton blend or tri-blend with less than 65% polyester, sublimation dye will not bond to the cotton fraction and will wash out, leaving dull, faded color.”
The rating stands; one line of our reasoning was wrong. We wrote that Sawgrass "never gives a percentage". The glossary we cited does not, but Sawgrass's Ink Sets page names polycotton 65/35 for its Sublijet UHD ink. That blend sits at SanMar's 65% floor, not below it, so the maker and the decorator guide agree. SanMar is not the only source that names the number; it is still the only one here that states the wash-out below it.
care.sawgrassink.com → - held ✓ Confirmed“Image Armor E-Series DTG inks cure white ink in 35 seconds, so a single 330°F-for-90-seconds rule is not the white-ink cure spec for every DTG ink.”
The rating stands; the DuPont figure in our reasoning was not DuPont's. We wrote that Firebird's page shows DuPont "requiring 180 seconds". Firebird's product copy tells buyers to say "goodbye" to waiting 180 seconds without naming anyone; the DuPont name comes from customer reviews on Firebird's own product page. DuPont's own process guidelines cure Artistri Brite P5000 on dark or black cotton for 60 to 75 seconds at 356°F (180°C), or 35 seconds at 365°F with a non-stick sheet. That is a third maker with its own time and temperature, so the conclusion that no single 330°F-for-90-seconds rule covers white ink holds. The 356F for Image Armor is correct as published.
dupont.com → - held ✓ Confirmed“The Kornit Apollo prints up to 400 pieces per hour, with wet-on-wet pretreat and inline drying.”
The rating holds — Kornit's technical specifications do state the claim. What did not hold was our caveat. We published “Impressions, not pieces”, and the same page says “Up to 400 shirts per hour” in its spec table and “350-400 prints per hour” in its FAQ. The distinction we drew is not one the source draws; the vendor contradicts itself, which is the stronger and more useful finding. The FAQ's own realistic range, 270-330, is now on the page too.
kornit.com →