Corrections
When something published here turns out to be wrong, we fix the page and write the fix down. A corrected page is not quietly re-edited — the change is dated and described here, so you can see what we got wrong and what it says now.
What is listed here
Pages that were already published: a factual claim that was wrong, a source that did not support what we said it did, a measurement that did not match the recipe, or structured data that told search engines something the page itself does not say.
Errors caught before publication are not listed. Most are — that is what the checks described in our Editorial Guidelines are for — and they never reach a reader. This log begins in July 2026, when we started keeping it.
Tell us about an error
Write to editor@tastekoreanfood.com with the page address and what you found. Reports that identify a real error are acted on, and health-related corrections are treated as urgent.
12 August 2026 — Recipe times sent to search engines (3 recipes)
What was wrong. The times printed on the page were right, but the structured data behind three recipes disagreed with them. Mango Bingsu was published to search engines as a twenty-minute recipe when the page itself calls for eight hours of freezing. Hwachae carried a twenty-minute cook time and five servings on a no-cook, fifteen-minute, four-serving recipe. Yellow Plum Syrup had no total time at all.
What it says now. Each one was rebuilt from the table readers see: eight hours twenty minutes, fifteen minutes with no cooking, and one hundred days of steeping.
How it was caught. A check that compares every recipe's visible time table against its own structured data, written after the first mismatch was noticed and then run across the site. Older recipes that do not use the table layout are not covered by that check yet.
12 August 2026 — Notes meant for our own editors were visible (16 recipes)
What was wrong. Sixteen recipe pages carried a sentence written for us, not for you — “Servings estimated from standard home versions — please confirm against your own recipe card before publishing.” and variations of it. It was an instruction to ourselves that should never have been published.
What it says now. The sentences are gone from all sixteen pages.
How it was caught. An outside review of the site. Before removing the sentences we checked the numbers they pointed at against each recipe's structured data — which is how the three timing errors above surfaced. A check now blocks that kind of sentence before publication, and the live site is swept for it daily.
10 August 2026 — A blood-pressure figure belonged to part of the trial, not all of it
What was wrong. Korean Black Garlic said a trial of 67 adults found diastolic blood pressure 5.85 mmHg below placebo. The number is real, but it is the result for the 32 men in that trial. Across all 67 participants the difference was 3.62 mmHg, and among the 35 women it was not statistically significant.
What it says now. The whole-group figure, with the split by sex described rather than hidden.
How it was caught. Our own re-reading of the source the day after publication. The paper's abstract leads with the subgroup number, so checking that the citation exists, that the journal and year match, and that the title fits the claim all pass — only opening the tables separates them. We now record which table a number came from.
17 July 2026 — A citation one digit off, and a claim stronger than its evidence
What was wrong. Omija (Schisandra Berry) cited a study ID one digit away from the paper we meant — 36558010 instead of 36558009, two papers published in the same journal on the same day — and took from it a percentage range that paper does not support. A second study cited on the same page was credited to the wrong journal, and the page did not say that its work was done in mice.
What it says now. The correct paper, the correct journal for the second one, the animal model stated plainly, the unverifiable figures removed, and a line saying that human skin trials are still lacking.
How it was caught. A citation audit that compares the title and subject of the paper against the claim it is attached to. Checking that a link resolves cannot catch this: both papers exist, in the same journal, on the same day.