FitChef

The log: one day.

May 25, 2026

2 corrections on 13 pages. Most arrive together: one paper read again against every page that cites it, and every page fixed the same day.

Six participant counts in the guides' evidence sidebars did not come from the studies and were wrong

May 25, 2026 · We got it wrong

What it said
The evidence sidebar on three guides showed participant counts that did not come from the studies' verified extraction data. In the supplements guide: 2,400 for a caffeine meta-analysis of 149 people, 814 for a BCAA review of 511, 3,600 for a creatine meta-analysis of 3,655, and 4,937 for a study of 4,755. In the protein guide: 24 for a study of 108. In the carbs guide: 1,779 for a glycaemic-index meta-analysis of 2,344.
What it says now
All six sidebar counts match the verified extraction data for each study, and each guide's sidebar total equals the total in its article text.
Why
The agent that builds the sidebar had no mandated source for participant counts, so it filled them in without one. The article text on every guide was checked and found correct; the error was confined to the sidebar layer. A rule now requires every sidebar count to come from the same grounded extraction files that feed the article (the linked system entry).
Built on it

25 shorts · 10 recipe hubs

Found by
Our own checks of the site.
Seven study pages could show a study count, or zero, as their participant count, and three claim pages carried wrong participant totals

May 25, 2026 · We got it wrong

What it said
On seven study pages the participant count was stored as free text (for example '49 studies, 1863 participants' or 'n=20'), and the page displayed the first number it found, so a reader could see a study count, or zero, where the participant count belongs. On three claim pages the total-participant figure was wrong.
What it says now
Each study page shows its verified participant count (for example 1,863 for the Morton 2018 protein meta-analysis, 3,877 for the viscous-fibre meta-analysis, 2,359 for the fat-burner meta-analysis, 20 for the fasted-cardio trial and 511 for the BCAA review). The three claims show 116, 303 and 11,113 participants, each checked against the studies they cite. One of the seven, the light-versus-heavy-load meta-analysis, was set to 684 here and later found to be unverifiable; see the record of 2026-06-03.
Why
A number-consistency audit found the rendering system silently misreading free-text sample sizes. Each value was re-checked against the study's grounded extraction before being written as a clean number.
Built on it

11 claims · 5 guides · 157 shorts · 321 recipes · 70 recipe hubs

Found by
Our own checks of the site.

Newer day: June 2, 2026The whole log

The full picture

How every claim on FitChef is checked, page by page.

We did the thinking. You just eat.

Get my week

It takes a minute.

ListenListen0:00Help