I get things wrong. When I do, the fix goes on the article where the claim was, and the record goes here with the date, so nobody has to trust that it was quietly cleaned up.
How to read this page
- ×The claim as it was published, in this color, followed by what changed.
- FixedThe page now says something different. The entry says what.
- OpenKnown to be wrong. Not rewritten yet.
- Reader nameWhen a reader catches it before I do, the credit goes here.
If you catch something before I do, send it to me. Tell me the page, the sentence, and where I can read the source you are looking at. If it holds up, the fix goes live and your name goes next to it here, if you want it there.
Most of the entries below came from one sweep in August 2026, when the whole Food Literacy Path and every ban claim on the Dirty List were checked against their primary sources. That is why the dates cluster. Formatting fixes and typos are not listed. Everything below changed what the site was telling you.
Corrected
Your Body Replaces 330 Billion Cells a Day. Protein Is the Raw Material. (Food Literacy Part 5)
- “Complete proteins contain all 9 essential amino acids in sufficient amounts. Sources: quinoa, hemp seeds, buckwheat, eggs, or combining legumes with grains.” Hemp seeds do carry all nine, but not in sufficient amounts: lysine was the limiting amino acid in every hemp product House and colleagues tested (Journal of Agricultural and Food Chemistry, 2010), and my own protein powder article already said so. Quinoa’s measured protein quality is moderate, which its Hub profile already said. Buckwheat came off because I had not checked it. The list now names fish and eggs, and the next line explains where hemp and quinoa sit.
Caught by me in an audit of every food statement on the site, 21 October 2026.
Broccoli, Health Benefits
- “A good plant-based source of protein, with 2.6 grams per cup, along with calcium for bone support.” 2.6 grams is about 5 percent of the 50 gram Daily Value. A good source means 10 percent or more, the bar the Nutrient Finder states. The line now says broccoli carries some protein.
Caught by me in an audit of every food statement on the site, 21 October 2026.
Swap Finder, staple swaps, white rice
- “Quinoa. Around 5g fiber and 8g protein per cup, and a complete protein.” Quinoa carries all nine essential amino acids and more lysine than rice or wheat, but its measured protein quality is moderate, as its own Hub profile says. The 8 grams stands. The swap now says “with all nine essential amino acids and more lysine than rice or wheat.”
Caught by me in an audit of every food statement on the site, 21 October 2026.
The Nutrition Hub, the glycemic panel on all 51 food profiles
- The Glycemic Scale Rating pointer. On 50 of 51 profiles the pointer sat at 1 to 5 on a scale printed 0 to 100. Those numbers were left over from an older rating scheme and were not glycemic index values. Sweet potato showed 3, where the published measurements for baked sweet potato run from 82 to 94. Oats showed 2, where the published value is 59. The pointer now appears only on profiles whose number I checked against a published measurement of that food in that form. Twelve profiles have one: banana 47, chickpeas 35, grapes 54, lentils 16, mango 48, oats 59, orange 45, papaya 38, pear 33, quinoa 52, strawberry 40 and sweet potato 87.
- The other 39 profiles. Each now says so in words under the scale. Twenty-nine are foods with too little carbohydrate for a glycemic index to be measured, such as arugula, garlic and the herbs. Ten have no number listed yet. Eight of those have no published measurement for the form the profile describes. Kiwi is one: its profile showed 58, which comes from one small test, beside a “Low” label that 58 does not earn. The other two, pineapple and watermelon, have published values that disagree with each other.
- “Low” labels with no measurement behind them. Nine profiles carried one: black radish, cherry, fennel seed, ginkgo, kiwi, milk thistle, mint, parsley and sage. Removed.
- The colored bands did not match the ranges printed on them. The green band ended at 50 on a computer and at 45 on a phone while it was labeled 0 to 55, so a correct low value could sit on yellow. The bands now follow the printed ranges.
- Still wrong, and waiting on a template edit. The line under “Glycemic Load (GL)” is labeled “GI per 100g”, which is neither a glycemic index nor per 100 grams. It is blank on most profiles and shows a stray number on a few. The glycemic load figures are worked out and go in when that label is fixed.
Caught by me in a review of the Hub’s glycemic data that began on 15 September 2026. The values come from the University of Sydney’s glycemic index database, using only tests run by the standard method in healthy people, checked where I could against the studies they cite.
The Dirty List, refined seed oils (now “Refined Cooking Oils”)
- The entry “Refined Seed Oils (Soybean, Corn, Canola)” has been rewritten and renamed. Every entry on the list has to rest on a named assessment or named studies that show harm. This one argued that the omega-6 fat in these oils drives inflammation and disease, and the sources it named do not show that. It is now about the one refining concern that does have a named assessment behind it.
- “Excess omega-6 promotes pro-inflammatory eicosanoid production, contributing to cardiovascular disease, obesity, and autoimmune conditions.” I wrote that as a finding. It is the argument of one review (Simopoulos, Biomedicine & Pharmacotherapy, 2002). A pooled analysis of 30 randomized trials found that eating more linoleic acid did not significantly change inflammatory markers in blood (Su and colleagues, Food & Function, 2017), and I left it out.
- “A 2018 study in the BMJ … found that replacing saturated fat with linoleic acid (from seed oils) did not reduce cardiovascular mortality and may have increased it.” The paper is from 2016 (Ramsden and colleagues). It re-analyzed one trial of corn oil and corn oil margarine and found no mortality benefit, which is not the same as finding harm. The American Heart Association’s 2017 advisory concluded the opposite about replacing saturated fat with polyunsaturated fat, and the entry did not mention it.
- “Shown in a 2020 UC Riverside study to cause neurological gene dysregulation and metabolic dysfunction in mice.” Male mice on high-fat diets, and the authors proposed that linoleic acid was not responsible for the hormone effect they measured (Deol and colleagues, Endocrinology, 2020). The source list also said the mice became obese. The paper reports a link with glucose intolerance, not with body weight.
- “Butter or ghee from grass-fed sources” and “Tallow or lard” as the healthier swaps. No source on the page compared those fats with the oils, and they run against this site’s own habit of leading with plants.
- “Lands (2005, Journal of the American Oil Chemists’ Society).” I could not match this to a paper. The Lands 2005 paper indexed in PubMed is in a different journal.
- 21 October 2026, follow-up. This site’s food stance changed on this date: whole foods still come first, and the cleanest fish and meat now have a place in small amounts. The fix above stands on its other grounds; its reference to leading with plants describes the stance at the time.
Caught by me in the same audit, finished on 12 September 2026, and the old entry stayed up until this date. The entry now covers the contaminants that refining leaves in oil (glycidyl esters and 3-MCPD esters), which the European Food Safety Authority assessed in 2016 and 2018 and the EU has capped by law. They are highest in palm oil, which is why palm is now in the title, and the entry says in its own text that it makes no claim about omega-6 fat.
The Dirty List, titanium dioxide (E171)
- “Look for: ‘titanium dioxide’ on labels, often buried mid-list.” Incomplete in a way that could mislead. The FDA does not require it to be named and lets it be declared as “artificial color,” so a label without the words can still belong to a product that contains it. The entry now says so.
- “…enough uncertainty that many people choose to avoid it, especially for kids’ snack foods.” I had no source for what many people choose, or for a separate risk to children. Removed. What the entry gives in its place is the record: the European Food Safety Authority could not rule out DNA damage (2021) and the EU took it out of food (in force for sales from 7 August 2022), while the joint FAO/WHO food additives committee (2023), the UK and the FDA weighed the same evidence and kept it. The FDA has had a petition to withdraw it since 14 April 2023 and, on its list dated 12 May 2026, still had it under review.
Caught by me in the same audit, finished on 12 September 2026. The same fix removed a line left in from drafting, “Here’s the grounded version:”, that sat at the top of the health section. It changed nothing the entry claimed, it should not have been published, and it stayed up from the day it was found, 12 September 2026, until this fix.
The Dirty List, TBHQ
- “At higher doses in animal studies, it caused stomach tumors and DNA damage.” On its own it did not cause tumors. The US National Toxicology Program fed TBHQ to rats for up to 30 months and to mice for 2 years (Technical Report 459, 1997) and found “no evidence of carcinogenic activity” in male rats, female rats, male mice or female mice. The nearest thing in the record is a set of experiments in which rats were first given known carcinogens and then TBHQ at 0.7 to 2 percent of the diet: in one, benign growths in the esophagus and forestomach went up and colon cancers went down. The European Food Safety Authority weighed those in 2004 and concluded that “TBHQ is not carcinogenic.” The DNA damage was in cell tests, and the one animal test that opinion calls conclusive was negative. The entry now reports all of that.
- “The National Library of Medicine notes acute symptoms such as nausea, vomiting, and ringing in the ears, but only at a 1 gram dose.” That passage is in the Library’s record for hydroquinone, a different chemical. Its record for TBHQ does not contain it. Removed.
- “A 2020 study from the Environmental Working Group found TBHQ in over 1,250 processed foods.” The number comes from the group’s news release of March 2021, which said “almost 1,250 other” foods. It is not in the group’s published paper. Removed.
- The source list. “Gharavi et al. (2007, International Journal of Molecular Sciences)” is a review in Current Drug Metabolism, not an immune cell study. “Butchko & Voelker (1997, Food and Chemical Toxicology)” matches nothing in PubMed. And I cited the EFSA opinion for “tumor promotion” without its conclusion, which was that TBHQ is not carcinogenic. The list now holds only documents I opened, with links.
Caught by me in an audit of the thirteen Dirty List entries written before August 2026, finished on 12 September 2026. The tumor line stayed up from that day until this fix, which is longer than it should have taken. TBHQ stays on the list for a different reason than the one it was listed for: both the joint FAO/WHO food additives committee and EFSA have shown how intake can pass the allowed daily amount when many foods carry the maximum, and two laboratory studies from Michigan State University report effects on immune cells and on allergy in mice. The entry says plainly that none of that is a study in people.
- A search for “ham” returned the formaldehyde entry, because “ham” sits inside “shampoos”. The Swap Finder matched a search word anywhere inside another word, and inside the list of alternatives. It now matches whole words in entry titles and listed examples first, and each result says what matched. A result is a lead to read, not confirmation of what is in a particular product. The first proposed fix missed half-typed and everyday words, so I held it back until those searches also passed review.
Caught by me while auditing both label tools on 16 September 2026.
- A label reading “BHA-free” was reported as containing BHA. The tool looked for ingredient names without reading the word in front of them, so a package advertising what it leaves out was read as a package listing it. The same thing happened with “no artificial colors”, “without BHA” and “no added corn syrup”. That is backwards on exactly the labels a careful shopper is most likely to check.
- I rejected the first fix because it hid ingredients that were actually listed. The revised matcher keeps punctuation while checking nearby absence wording and reads line breaks as wrapped text. It now skips simple phrases such as “BHA-free”, for butylated hydroxyanisole, and “no added corn syrup”. It keeps butylated hydroxytoluene (BHT) visible in “Contains BHT. Free of artificial dyes.” It also keeps a wrapped “high fructose corn syrup” together. One shape is still read wrong: after a single “no”, only the first name in a list is skipped, so “no BHA or BHT” still reports BHT. This is still a limited name check, not a reading of every possible sentence. I checked the revised behavior on the live page before marking this item Fixed.
Caught by me while auditing both label tools on 16 September 2026.
Diet Discipline Coach: Privacy Policy (now version 1.4)
- “Deleted data can survive in the hosting provider’s automatic backups for a short window, currently about 7 days…” The database plan the app runs on keeps no restorable backups, so this described copies that did not exist. I first saw that on 12 September 2026, when a check of the database plan found no restorable backups, and I left the sentence up while I decided how to reword it. That was a mistake. The copies that do exist are my own: the app’s database is copied before I change how it is built or repair data in it, and each copy is kept encrypted on my own computer and deleted once it is 6 days old. Section 7 now says exactly that, including the one case where a copy can last longer.
- “If a timed purge is ever added, … it will be described here before it runs.” Those copies, and the timer that deletes them, started on the evening of 7 October 2026, a day before this page described them. That broke the page’s own promise, and it made “deleting your account removes it in one step” untrue while a copy existed. Sections 5, 7, 8 and 9 now describe the copies.
- “No cookies. This is a native app, not a website.” The web version is a website. What is true, and what the line now says, is that the app’s own code sets no cookies; section 3.3 lists what a browser keeps.
- “It is invitation only and there is no public sign up, so a minor cannot create an account here.” Anyone can create a login, and until 8 October 2026 a login without an invitation could also set up a coaching workspace of its own. That was closed on 8 October. The page now says what is true: without an invitation, a login cannot see anyone else’s data, and the app offers it nothing but redeeming an invitation, signing out and deleting itself.
- “When you delete your account they are anonymised…” Records of actions about a deleted student keep their record numbers and, where I archived a student, the note I wrote. That is not anonymous. Section 7 now lists what stays.
- “No one else. There are no other processors, no subcontractors…” Broader than I can stand behind. Section 5 now says only what is true: no data brokers, no advertising partners, and your data is never sold or shared for anyone else’s marketing.
Caught by me on 8 October 2026, in an audit of the app’s legal pages and launch steps before the first students join. Version 1.3, live for under an hour that evening, still said more than was true in three places (a revision note in the audit records, what a login without an invitation can do, and what I can see of private entries). Version 1.4 fixed them the same night.
Diet Discipline Coach: Terms of Use (now version 1.2)
- “It is invitation only. There is no public sign up.” Anyone can create a login. The terms now say what is true: without an invitation, a login cannot see anyone else’s data, and the app offers it nothing but redeeming an invitation, signing out and deleting itself.
- “These terms cover the Diet Discipline Coach mobile app…” Since 13 September 2026 the app also runs in a browser, and that is now the main way to use it. The terms now cover the web version and the Android app.
Caught by me on 8 October 2026, in the same audit. Version 1.1, live for under an hour, still overstated the sign-up line; version 1.2 fixed it the same night.
The homepage (the four running totals near the top)
- “4 Core Tools.” There are three: the Swap Finder, the Label Decoder and Find a Food by Nutrient, which is what the Tools page itself says. The number was typed in by hand, and it was wrong. It now says 3.
- “14 Articles Published.” There are 22. That number was also typed in by hand and was not updated as articles went up. It now says 22. The Nutrition Hub total, which said “49+”, now gives the exact count, 51.
Caught by me in an audit of the whole site on 2 September 2026. The homepage was not fixed until 7 October 2026, which is longer than it should have taken.
Diet Discipline Coach: Privacy Policy (now version 1.2)
- “Relays push notifications to your phone.” That was the whole of what the policy said Expo does, and it was incomplete. Since the web version of the app first went up on 13 September 2026, Expo has also hosted it, and like any web host it receives your IP address and browser type when you open it. What you write in the app goes from your browser straight to the database, never through Expo. Section 5 now says so.
Caught by me on 4 October 2026, while checking the policy against the app’s code before extending it to cover the web version.
Diet Discipline Coach: Privacy Policy (now version 1.1)
- “Your push token and the generic text of a notification.” That was the full list of what Expo and Google receive when the app sends a push, and it was incomplete. Every push also carries routing details the app uses to open the right screen: the kind of notification, internal record numbers, and for some kinds a count, the days overdue or whether homework was approved or sent back. None of it is anyone’s content, and pushes have carried it since they were built in August. Section 5 now lists all of it.
- “A reminder says ‘one tracker is still open today’.” The app never sent that text. The tracker reminder says “A tracker is waiting for an entry.” The example is fixed in the short version and in section 6.
- “At most one of each kind per day, overdue reminders stop after three days.” Not exact. Task and tracker reminders come once for each reminder time you chose, so they can arrive more than once a day, and overdue reminders stop after the number of days set for that homework, three by default. Section 6 now states the real limits.
Caught by me on 2 and 3 October 2026, while checking every notification the app can send against the database code that creates it, before the new private message alert went live.
Meal Prep Is Not About Tupperware. It Is About Removing Decisions. (Food Literacy Part 16)
- “Every day, you make an estimated 200+ food-related decisions, according to research by Wansink and Sobal.” The figure comes from a 2007 paper that its publisher, SAGE, placed under an Expression of Concern on 17 March 2026. A number its own publisher has flagged should not carry an argument, so it is gone from the article and from its summary box.
- A “What to Watch Over 2-4 Weeks” list of what readers would notice. It was framed as what “the patterns suggest”, and no named study was behind it. The list is gone.
Caught by me on 25 September 2026, in a sweep of the Food Literacy Path for claims that lean on research without naming it. Each fix was checked against the paper itself.
The Eating-Speed Trap (Food Literacy Part 15)
- “A 2021 meta-analysis in Frontiers in Nutrition confirmed the pattern across multiple populations: fast eating was associated with a 54% higher risk of metabolic syndrome…” Those figures are odds ratios, not risk, and 28 of the 29 studies behind them were observational, so they show an association rather than confirm a cause. The page now says so and names the paper (Yuan et al., 2021).
- “Fast eating raises blood pressure, blood sugar, and waist circumference…” and “the benefits showed up when people actually changed it.” Both described observational results as causes. Both now say what the studies measured and that they were observational.
- A “What to Watch Over 2-4 Weeks” list presented as what the research suggests. No named study was behind it. The list is gone.
Caught by me on 25 September 2026, in a sweep of the Food Literacy Path for claims that lean on research without naming it. Each fix was checked against the paper itself.
Why One Superfood Won’t Save You, But a Pattern Will (Food Literacy Part 14)
- A retracted paper was cited. The article cited Estruch et al. (New England Journal of Medicine, 2013) for lower CRP and IL-6. That paper was retracted and republished in 2018, and it reported heart attacks and strokes, not those markers. The PREDIMED substudy that did measure them (Mena et al., 2009) found IL-6 fell on both Mediterranean diets but CRP fell only on the olive oil version, and the page now says exactly that.
- “Umbrella-level evidence suggests omega-3 intake may reduce CRP, IL-6, and TNF-α in some populations, with stronger effects often seen when baseline inflammation is higher.” The review behind it pooled supplement trials, not food, and its summary reports no baseline pattern. The page now says supplements, and the second half is gone.
Caught by me on 25 September 2026, in a sweep of the Food Literacy Path for claims that lean on research without naming it. Each fix was checked against the paper itself.
What You Ate Today Is Deciding How You Sleep Tonight (Food Literacy Part 13)
- “A single day of low-fiber, high-fat, high-sugar eating was enough to measurably degrade sleep quality that same night.” The St-Onge study (2016) fed 26 adults a controlled diet for four days, then let them choose their own food on the fifth. The diet and sleep links come from that one self-chosen day, so they are associations, not a measured effect of a bad day of eating. The page now says that, and it no longer credits saturated fat with more night-time arousals; that finding was for sugar.
- “Eating a large meal within 2-3 hours of bedtime is consistently associated with poorer sleep.” The one study named for it (Crispim et al., 2011) was a single observational study of 52 adults, and its fat and calorie findings held in the women. The page now says exactly that.
- “Falling asleep within 15-20 minutes of lights out, rather than 30-45.” It sat in a “What to Watch” list presented as what the research suggests, and no study behind the article measured it. The list is gone.
- Two unsourced paragraphs on meal timing. One gave a body-temperature mechanism as settled, and the other suggested a small snack before bed. The only study the article cites on timing (Crispim et al., 2011) found eating close to bedtime, late-night snacks included, associated with poorer sleep. Both paragraphs are gone.
Caught by me on 25 September 2026, in a sweep of the Food Literacy Path for claims that lean on research without naming it. Each fix was checked against the paper itself.
You Could Eat Enough Calories and Still Be Malnourished (Food Literacy Part 8)
- “Regular adequate intake may shorten cold duration slightly.” The Cochrane review behind it tested supplements of at least 200 mg a day, not ordinary intake from food. It found colds 8% shorter in adults and 14% shorter in children, and no fewer colds in the general population. The page now says exactly that.
Caught by me on 25 September 2026, in a sweep of the Food Literacy Path for claims that lean on research without naming it. Each fix was checked against the paper itself.
Your Body Replaces 330 Billion Cells a Day (Food Literacy Part 5)
- The protein table credited to Phillips (2017). It gave 0.8-1.0, 1.2-1.6, 1.0-1.2 and 1.6-2.2 g/kg/day for sedentary, active, older and athletic adults. Phillips does not give those ranges for those groups: he puts a safe minimum at 1.0-1.2 for healthy younger adults and at least 1.2 for older adults. The table now shows only figures its sources give, from Phillips, the PROT-AGE group (Bauer et al., 2013) and Morton et al. (2018). The source list also had the wrong journal for Phillips.
Caught by me on 25 September 2026, in a sweep of the Food Literacy Path for claims that lean on research without naming it. Each fix was checked against the paper itself.
You’re Probably 10-20 Grams Short on Fiber (Food Literacy Part 9)
- “Viscous soluble fiber has shown meaningful LDL reduction in meta-analysis data.” The meta-analysis behind it (Brown et al., 1999) found the effect small: about 5 mg/dL of LDL for 3 grams of soluble fiber from oats, and its authors say soluble fiber can make only a small contribution. The page now gives the size.
- “Higher fiber intake is linked with lower cardiovascular risk in large pooled data sets, with stronger protection as intake rises.” The pooled studies were observational, so they show an association, not protection. The page now says 9% lower risk per extra 7 grams a day, in observational studies.
Caught by me on 25 September 2026, in a sweep of the Food Literacy Path for claims that lean on research without naming it. Each fix was checked against the paper itself.
Your Brain Is 60% Fat (Food Literacy Part 4)
- “Replacing it with unsaturated fat improves outcomes. Replacing it with sugar does not.” That is the American Heart Association’s reading (Sacks et al., 2017), and the page presented it as settled. The independent Cochrane review (Hooper et al., 2020) found a smaller benefit from cutting saturated fat, little or no effect on cardiovascular deaths, and no clear difference between replacing it with polyunsaturated fat or with carbohydrate. The page now names both.
- “Omega-3 fatty acids (EPA and DHA from fish; ALA from plants) are anti-inflammatory, support brain function, and are associated with reduced cardiovascular risk.” The review cited for it (Mozaffarian and Wu, 2011) is about heart disease, not the brain, and it says omega-3s might lower inflammation, not that they do. The page now says what that review found: seafood omega-3s lower triglycerides, resting heart rate and blood pressure, with the most consistent benefit for deaths from coronary heart disease.
Caught by me on 25 September 2026, in a sweep of the Food Literacy Path for claims that lean on research without naming it. Each fix was checked against the paper itself.
48 of 54 Avocado Oil Samples Failed. The Olive Oil on the Same Shelf Passed., the credit for where the article came from
- It ran with no credit to the person who got there first. I wrote it after watching I Read Six Avocado Oil Studies. It’s Worse Than You Were Told. by Cassandra at Becoming a Farm Girl. The argument is hers: setting the olive oil result beside the avocado one, and turning that into something a shopper can actually check. I read the studies, the statute and the standards against the primary documents myself, so the sourcing is my own work. The credit was not, and it should have been on the page from the first minute. It was missing for about eleven hours.
- Several sentences were hers word for word and were not marked as hers. The suggested email to a brand, which ran as an unattributed quotation; “a claim about a bottle is not the same claim about the mayonnaise”; and “it is usually the cheapest way to be right.” Those three are now credited in place. A few shorter phrasings are hers too, and the note at the top of the article now says so rather than marking each one.
Cherry, Zucchini, Kiwi and Brussels sprouts, the nutrition numbers printed under the serving size
- Cherry: “Calories: 63, Sugar: 12.8g, Fiber: 2g, Protein: 1g” under a serving of 1 cup pitted (154g). Those are the figures for 100 grams, not for the cup on the label. A cup of sweet cherries is 97 calories, 19.7g of sugar, 3.2g of fiber and 1.6g of protein. Every number understated the food by about a third.
- Zucchini: “Calories: 17, Protein: 1.2g, Fiber: 1g, Sugar: 2.5g” under a serving of 1 medium (196g). The same error, on all five figures. One medium zucchini is 33 calories, 2.4g of protein, 2g of fiber and 4.9g of sugar.
- Kiwi: “Calories: 61, Sugar: 9g, Protein: 1.1g” under a serving of 2 medium kiwi (138g). The same error again. Two kiwis are 84 calories, 12.4g of sugar and 1.6g of protein.
- Kiwi: “Fiber: 14.7g”. A separate mistake and a worse one. Two kiwis carry about 4.1 grams of fiber. As published, it told you that two pieces of fruit delivered more than half a day’s fiber.
- Brussels sprouts: a serving labelled “1 cup raw (156g)” carrying cooked calories and raw vitamins. The calories, protein, fat, fiber and sugar all matched cooked sprouts exactly. The vitamin C and vitamin K percentages all matched raw sprouts exactly. And 156 grams is the weight of a cup of cooked sprouts, where a cup of raw is 88. One entry was describing two different foods at once. It now reads 1 cup cooked (156g) throughout, with the vitamins and minerals recomputed on cooked: vitamin K 182% rather than 230%, vitamin C 107% rather than 147%.
Caught by me on 9 September 2026, checking every number field in all 51 published Hub entries against the USDA FoodData Central SR Legacy tables. Of the 48 that could be checked, 33 were already correct.
Arugula, Brussels sprouts and Red cabbage, the first three entries rewritten to the standard the rest of the site already meets
- “Sulfur compounds associated with the body’s natural detoxification processes.” Said on all three entries in slightly different words. No human study supports it, and it is the exact phrasing this site refuses everywhere else. Arugula now explains what glucosinolates actually do when a leaf is cut, chemistry you can taste. Brussels sprouts names the two small Dutch feeding trials that measured an enzyme change and says plainly that an enzyme change is not a health outcome. Red cabbage now states composition only.
- “Contains nitrates similar to beets, which the body may convert to nitric oxide to support healthy blood flow.” (Arugula.) The evidence is split, and the entry now says so: a Danish cohort, the Nurses’ Health Study and two feeding trials, all named, two pointing one way and two the other, and none of them a trial of a serving of arugula. No blood-flow benefit is claimed.
- “A source of calcium from a plant-based food, contributing to bone mineral density.” (Arugula.) A serving carries about 2.5% of the daily value. Cut.
- “US rules require vitamin C in bacon.” (Arugula, in the note that points to the Dirty List nitrate entry.) The rule, 9 CFR 424.22, requires sodium ascorbate or sodium erythorbate in pumped bacon, which is narrower than I wrote. The note now states the rule as written and sources the chemistry to a named 1991 review instead of asserting it.
- “High in fiber and protein for a vegetable, promoting satiety and supporting digestive health.” (Brussels sprouts.) A good source of fiber with a little protein, and no human study of Brussels sprouts and satiety exists to cite. The entry now states the fiber’s established role and nothing more.
- “Contains kaempferol, an antioxidant flavonoid associated with reduced inflammation in research studies.” (Brussels sprouts.) No human study of Brussels sprouts, or of kaempferol at the dose a serving carries, supports it. Cut.
- “Anthocyanins, potent antioxidants associated with reduced inflammation.” (Red cabbage.) The entry now names the one feeding study that shows the pigments are taken up, in tiny recovered amounts, and the one four-week trial that fed red cabbage and found no statistically significant change in immune or antioxidant markers. No health claim is made.
- “Over 270% of the daily value” (vitamin K), “over 160%” (vitamin C), “85% of the daily value” (vitamin C). (Brussels sprouts and red cabbage.) These percentages were computed on the daily values that were retired in 2020; on the current ones they are about 230%, 147% and 56%. The nutrient fields on the page carry the current figures, so the rewritten prose no longer repeats percentages at all.
Caught by me in the sweep of 3 September 2026 recorded below. Every claim that survived now names its study, population and dose; the rest were cut, and the searches that came up empty are on file. Posted 8 September 2026.
- “One cup delivers 8.1g of protein, 5.2g of fiber and over half a day’s manganese, with all nine essential amino acids present.” True as written, and misleading, which is worse. Naming the nine amino acids implies complete, high quality protein. Measured scores say otherwise: published DIAAS for quinoa flour is about 50, and PDCAAS estimates run from roughly 0.57 to 0.85, none of them exceptional. Quinoa is also not unique in carrying all nine, despite a widely repeated line in a 2011 FAO report; buckwheat, amaranth and hemp do too. The entry now says quinoa carries more lysine than rice or wheat, which is the real and checkable advantage, and states plainly that its measured protein quality is moderate.
Caught by me a few hours after publishing the entry, when international sources came back on a separate question.
A Cabbage Outscored the Blueberry. Here Is What “Super” Actually Measures.
- “Wang and colleagues (Circulation, 2021) followed more than 100,000 US adults for up to 30 years and pooled 26 other long-term studies.” The paper pooled its own two cohorts with 24 others, 26 in all. The page now says so.
- “Spinach, 86.43: Fresh or frozen.” The CDC nutrient density scoreboard scored raw foods only, and frozen spinach would score lower per calorie on the USDA numbers the same article cites. The note now reads “Scored raw, like everything here.”
- “Austria and colleagues gave people whole, milled or oil-form flaxseed for a month.” The trial ran three months, with the rise in blood levels measured in the first month, and the authors noted the whole-seed group struggled to keep eating the seeds. Both facts are now on the page.
- “Cooking inactivates the enzyme involved.” The source says cooking “largely deactivates” it. The hedge is restored.
Caught by me in a same-day audit of the article, about an hour after it went up.
Ginkgo biloba, turmeric and rosemary
- “Some research suggests modest short-term cognitive support in individuals with existing cognitive decline when used alongside conventional care.” A claim about a medical condition, and about using a supplement alongside medical care, carrying no source. The largest trial ever run on the question followed 3,069 adults aged 75 and over for about six years and found no reduction in dementia, with the rate slightly higher in the ginkgo group (DeKosky et al., Journal of the American Medical Association, 2008). The page now leads with that. The same entry described ginkgo’s compounds as having “circulation-supporting properties” in a second field, which is fixed too, because a claim can sit in more than one place on a page.
- 13 October 2026, follow-up. When I fixed this page in September I wrote that the circulation work was “mostly cell studies and small trials.” That undersold the record. A Cochrane review had already pooled 14 placebo-controlled trials in 739 people with leg pain from narrowed arteries and found no clinically significant benefit (Nicolaï et al., Cochrane Database of Systematic Reviews, 2013). The page now names the trials on both sides and says a circulation benefit is possible but not guaranteed.
- 13 October 2026, follow-up. The ginkgo page now carries a sourced sentence on dementia. A Cochrane review published in February 2026 pooled 13 placebo-controlled trials in 3,288 people who already had dementia and found there may be small to moderate benefits at six months, on low-certainty evidence (Wieland et al., Cochrane Database of Systematic Reviews, 2026). My September fix removed the unsourced line and did not mention that review. The trial at the top of the page still stands: ginkgo did not keep dementia from starting.
- “Research suggests it supports joint comfort and mobility” and “studies indicate potential benefits for brain function and memory.” Two condition claims about turmeric, neither naming a study. The page now explains the thing that actually matters: the research is done on concentrated curcumin extracts, and in humans given 2 grams of curcumin alone the blood levels were undetectable or very low (Shoba et al., Planta Medica, 1998).
- “Rosmarinic acid may support cognitive function, focus, and memory retention.” No large human trial supports it, and it is a claim about how a brain works. Gone. Rosemary is a tablespoon of flavor at 2 calories and the page says so.
Caught by me on 3 September 2026 in the first full sweep of the Nutrition Hub. All three pages now say on the page itself what they used to claim, because a site that corrects itself out loud is worth more than one that was never wrong.
Carbs aren’t the enemy (Food Literacy Part 3)
- “This slows digestion and prevents blood sugar spikes.” Prevents was too strong, and it is the absolute shape this site avoids everywhere else. Pairing carbohydrate with protein, fat or fiber lowers the rise and pushes it later. It does not reliably stop one. The page says that now.
- 21 October 2026, follow-up. This site’s food stance changed on this date: whole foods still come first, and the cleanest fish and meat now have a place in small amounts. The fix above stands on its other grounds; its reference to leading with plants describes the stance at the time.
Caught by me on 3 September 2026, by a scan run over the article straight after rewriting a different section of it. The same sentence used chicken as its example, which contradicts this site’s own habit of leading with plants, so it now reads rice with lentils.
Carbs aren’t the enemy (Food Literacy Part 3)
- “Your body breaks them all down into glucose.” Not all of it. What you digest becomes glucose, fructose and galactose, and fiber is not broken into glucose by human enzymes at all. The article now explains that split, because it is the whole reason two foods with the same carbohydrate number behave differently.
- Simple carbohydrates break down quickly, complex ones slowly. Too clean to be true. Some starches digest fast, and the real answer depends on processing, particle size, ripeness, cooking, cooling, the fiber and the rest of the meal. That split is gone, replaced with the one that holds up: how much was stripped out before the food reached you.
- “Symptoms anyone who’s tried an extreme low-carb diet has likely felt.” Some people get those symptoms and some do not. The page says that now.
Caught by me in the August accuracy audit. The replacement is built on four foods read from the USDA database that day, per 100 grams: raw spinach carries 3.6 grams of carbohydrate and 2.2 of it is fiber, while regular cola carries 10.4 grams and none of it is. Same line on a label, opposite thing in your body.
What food actually is (Part 1) and Carbs aren’t the enemy (Part 3)
- “Carbs aren’t optional extras” and “the question isn’t whether you need them.” Both said you have to eat carbohydrate. You do not. Your liver makes glucose from protein and from part of the fat molecule, so a person can live without eating any. Both pages say that now, and both carry the other half of it, which is that the Food and Nutrition Board still sets a recommended intake of 130 grams a day.
- “Your brain still needs glucose.” Too flat. When carbohydrate is scarce the liver also makes ketones and the brain burns those. Part 3 now says so and names the study, including the part that matters: it was five to six weeks of starvation, not a diet.
Caught by me in the August accuracy audit, and it was the last item from that audit still open. Checked against the Food and Nutrition Board’s own tables, where a recommended intake of 130 grams a day sits next to a recommendation that 45 to 65 percent of calories come from carbohydrate, which on a 2,000 calorie day is about 225 to 325 grams. Both numbers are on the page now, because printing one of them alone is how this went wrong the first time.
Thin on the Outside, Fat on the Inside
- Waist measurement. The article said to measure level with the navel. The published self-measurement method (NICE guideline NG246) puts the tape midway between the lowest rib and the top of the hip bone, which sits higher on most people. Fixed in both places it appeared, and the article now says a number taken at the navel is not comparable to one taken the right way.
- Vitamin D. It was listed among the blood markers that “see inside” for visceral fat. It is not one. Season, sun, skin tone and supplements all move it. Removed from that list.
- Clearing pollutants. The article told you to support “the routes your body already uses, which run on fiber, water and regular bowel movements.” That mechanism has no human evidence behind it, and it sat one paragraph after the article said there was no evidence for detox programs. Removed, and the article now says why.
- Vitamin D after sun exposure. The study cited measured the increase in blood levels, not the peak. Reworded to match what was measured.
Caught by me in the August accuracy audit.
Your Body Replaces 330 Billion Cells a Day (Food Literacy Part 5)
- “Your body can convert carbs to fat and vice versa.” Half true. Surplus carbohydrate does become fat, but the fatty acids in fat do not become glucose in any meaningful amount. Rewritten. The point that matters is nitrogen: protein carries it, carbohydrate and fat do not, which is why protein cannot be substituted.
Caught by me in the August accuracy audit.
The Nutrition Label Is Talking to You (Food Literacy Part 7)
- “Every claim on a food label is technically true. It has to be. The FDA requires it. The calorie count is accurate.” That was false, and it was the article’s thesis, stated twice. Rounding rules mean under 0.5 g of trans fat prints as 0 g, and a compliance tolerance lets a bar labelled 200 calories hold 240. Rewritten: a label is regulated, not exact.
- “‘Made with whole grains’ can mean 1% whole grain.” Stated with no source. Now says what the rule actually is: no minimum proportion is set, so the phrase does not tell you how much.
Caught by me in the August accuracy audit.
The Dirty List, azodicarbonamide (ADA)
- “Banned in the EU, UK, and Australia.” The EU’s food additive regulation (Regulation (EC) No 1333/2008) does not list it, which makes it not authorized, and that is a different thing from banned. The UK and Australia claims could not be traced to a primary source and were removed. They were not disproved, they were unsupported.
- 1 September 2026, follow-up. The summary block at the top of the Dirty List page still said “not authorized in the EU or UK.” The 27 August fix corrected the entry and missed the summary above it. “Or UK” is now removed there too.
Caught by me in a sweep of every ban claim on the Dirty List. 18 entries checked, 2 were wrong.
The Dirty List, BPA
- “The EU adopted rules restricting” BPA, with no citation. Now cites the regulation itself: Commission Regulation (EU) 2024/3190, Article 3(1), which prohibits BPA in food contact materials.
Caught by me in the same ban-claim sweep.
Apeel: What Is Really Coating Your Produce? (Food Literacy Part 6)
- Trace solvents and heavy metals “acknowledged” in the FDA filing. The seven substances came from a table of maximum permitted limits, not from measurements. A limit is not a detection. The claim was repeated in three places and removed from all three, with a note at each.
- The manufacturing process. The ethyl acetate, heptane and palladium process quoted in nearly every online critique was described as how the coating is made. FDA’s May 2024 letter records that process was a proof of concept that was never commercialized. Rewritten.
- “EU/UK: approved only for thick-skinned produce.” The EU authorizes the additive, not the brand, on a named list of fruits. It is a list, not a thickness rule. Kiwi is on it and apples are not. The UK was dropped from the row because its current version of the list could not be verified.
- The coating “bonds to the produce” and survives peeling. Not sourced. Replaced with FDA’s own wording, a thin edible physical barrier. Peeling removes the peel.
Caught by me in the August accuracy audit. The article carries a dated correction notice at the top and a note at each claim.
- All four were re-checked against their sources in the same audit. The one that changed a conclusion: Part 11 quoted a 2021 fiber and fermented food trial as 10 weeks. It ran 17 weeks with 18 people per arm, and its primary outcome was null. The article had quoted only the secondary finding. Both facts are now in the article.
Caught by me in the August accuracy audit.
Known and not yet fixed
Seven Hub entries whose numbers I could not confirm, or could not confirm against the right food
- Oats and sunflower seeds list vitamin and mineral percentages belonging to a different product than their calories do. The oats calories, protein and fiber match plain unfortified rolled oats, while the thiamin, manganese and copper percentages match a fortified product. Sunflower seeds list pantothenic acid at 39% of the daily value, which is the figure for dry roasted kernels, while the rest of the entry matches dried. I am not going to guess which product each entry means, so both stay flagged until I decide and rewrite them together.
- Turmeric’s numbers are for roughly a teaspoon, under a label that says 1 tablespoon ground (6.8g). Either the label or the numbers is wrong. I want to weigh a spoonful before I pick which.
- Kimchi, milk thistle, ginkgo biloba and dragon fruit cannot be checked against the USDA tables at all. Milk thistle and ginkgo are sold as extracts and teas, not as a food with a serving size. Dragon fruit is not in the database. The database’s own kimchi record is thin and lists zero vitamin C for a fermented cabbage, which I do not believe, so correcting the entry to match it would make the page worse rather than better. These four need a different named source, or they should stop showing numbers.
Caught by me in the same 9 September 2026 check. They are named here rather than quietly left alone, because an unverified number looks exactly like a verified one.
The Nutrition Hub, 47 of 50 entries still to do (arugula, Brussels sprouts and red cabbage were rewritten on 8 September 2026, see above)
- Not one Hub entry cites a source. I read all 50 end to end on 3 September 2026. There are 5,437 words of health claims in the Hub and none of them name a study, an author, a year or a body. Everywhere else on this site a claim has to be anchored to something you can go and check. The Hub was never held to that, and it is the largest set of claims here.
Caught by me in a sweep of the whole Hub on 3 September 2026, run because I had said here that the Hub was unchecked. It is checked now. The nutrient fields looked right where I checked them, and that full check against the USDA tables has now been done, on 9 September 2026, with both results recorded on this page; the percentages written into the prose, though, were on the pre-2020 daily values, which is why rewritten entries no longer repeat percentages. It is the benefit writing that has to change, and it changes entry by entry, in batches, so this stays open until all 50 are done.