Corrections

When we're wrong, we say so.

Every correction is published here, dated, in the open. That's the point.

• 2 November 2026 — Pulling an all-nighter: This page's summary of its evidence said "mixed designs" and "a link, not proof of cause". The reviews behind it are of sleep-deprivation experiments, and the summary now says so. No finding or number changed, and neither did the limits stated beside each result: most studies assigned people at random or in a balanced order, but not every one did.
• 2 November 2026 — Where you walk, not whether you walk: This page's summary of its evidence said "a link, not proof of cause" beside a review of experiments. The studies behind the headline are experiments, and the summary now says so. No finding or number changed, and neither did the limits stated beside each result. The separate comparison of greener and less green places is still observational and is still called a link.
• 2 November 2026 — Never downshifting out of stress: This page's summary of its evidence said "a link, not proof of cause" beside "randomised trials", which reads as a contradiction. These are randomised trials, and the summary now simply says so. No finding or number changed, and neither did the limits stated beside each result: people knew whether they were breathing slowly, so expectation may explain part of the benefit.
• 1 November 2026 — A couple of drinks a night: This page labelled the liver review as a meta-analysis of cohort studies. Of its 44 studies, 26 were cohort studies and 18 were case-control studies, so the label now says meta-analysis. The numbers are unchanged.
• 1 November 2026 — Headphone volume: This page labelled its sources as meta-analyses of cohort studies, and one as a cohort study. None is. The prevalence review accepted several designs without saying how many of each, the hearing-test review compared groups using case-control or cross-sectional studies, and the third review pooled nothing. The labels now say so. The page also said some device studies relied on what people reported about their own volume; that review required a measured output level, and differed instead in the limit it counted as unsafe. The numbers are unchanged.
• 1 November 2026 — "Running ruins your knees": From reading the anchor review in full. This page said the comparison group was sedentary people, not walkers. About 61% of the controls in the recreational comparison were walkers from one study, so the page now calls them non-runners and says so. It said the review gave no figure for more than 15 years of running and described a curve with two rising ends. The review gives 1.01 (0.72 to 1.43), no difference from controls, from studies that were all of competitive runners. It said the review gave no figures for the knee alone or by sex; those are now on the page. The overall result is unchanged.
• 1 November 2026 — "Running ruins your knees": This page labelled two of its reviews as meta-analyses of cohort studies. Neither is. By its own tables, the running-level review pooled 17 studies: 6 cohort, 7 case-control, 3 cross-sectional and 1 listed as both case-control and cross-sectional, so its label now says meta-analysis. The knee surgery figure came from 3 case-control studies and is now labelled that way. The numbers are unchanged.
• 1 November 2026 — Steps: This page said most of the benefit of walking has levelled off by about 7,000 steps and called that a sweet spot. That holds for only some outcomes. The review found the curve flattening around five to seven thousand steps for early death, heart disease, dementia and falls, and carrying on in a straight line for heart disease death, cancer, type 2 diabetes and depressive symptoms. The page now says there is no single target. The numbers are unchanged.
• 1 November 2026 — Phone in bed: This page said the headline figure for not getting enough sleep pooled twenty studies and over 125,000 children. Those are the totals for the whole review; that figure pooled seven of its twenty studies, and the review does not report how many children those seven covered. The figure itself is unchanged.
• 30 October 2026 — Cutting back on social media: Two results on this page, the one about taking a complete break from social media and the earlier digital-detox review, were labelled "Systematic review (no pooled estimate)". Both reviews did pool their studies into a single estimate, which is the figure shown; the studies they pooled were not all of one design, randomised and non-randomised together. They are now labelled "Meta-analysis". The numbers and the findings are unchanged.
• 13 October 2026 — Phone in bed: This page said the phone harms sleep by keeping you mentally wound up as well as by keeping you up later, and that scrolling in bed does the opposite of helping you wind down. A 2026 meta-analysis that followed the same young people from day to day found no significant link between screen time and how long it took them to fall asleep or how well they slept; the better-supported explanation is that the phone keeps you up later. The page now says so, and no longer says that the trial and the size of the association point the same way. The fix also used to include a screen-free last hour before bed. Its only support was an exploratory subgroup in one trial and a 38-person pilot, so the fix is now only to put the phone across the room. Two sentences were removed outright: one suggesting a standalone alarm clock, because a fix here never involves buying anything, and one saying blue-light glasses have the same problem as night mode, because no source on this page tested glasses. The numbers on this page are unchanged.
• 12 October 2026 — Movement after meals: This page said a walk after a meal works because leg muscles pull sugar out of the blood just as the meal is absorbed, and gave that as the reason the timing matters. That explanation is a hypothesis in the reviewers' discussion, not a measured finding, and the same reviewers consider other processes more likely. The page now says so. The findings themselves are unchanged: the lower blood-sugar rise, and walking after the meal beating the same walk before it.
• 8 October 2026 — Short sleep: The number of people behind the short-sleep figure was given as 1,382,999. That is the review's total across all of its analyses; the short-sleep analysis itself pooled 25 cohorts from 15 studies with 1,381,324 people and 112,163 deaths. The finding (about 12% higher risk, RR 1.12) is unchanged.

Report an error

If something here is wrong — a number, a source, or a conclusion the evidence does not support — tell us. A person reads this.

Email onehonestfix@gmail.com and include:

• Which page or video — a link, or the topic name.
• The specific claim — quote it if you can.
• Why you think it is wrong — a source is ideal, a reason is enough.

What happens next. The claim is checked against the paper it came from. If it is wrong, the source of truth is corrected, every page built from it is regenerated, and the correction appears on this page with the date and what changed. If a published video repeats the error, that is recorded here too — a posted video cannot be quietly edited, so the correction is the record. If the claim turns out to be right, you get an answer explaining why; the check is the same either way.

One person reads this inbox, so a reply may take a few days. Please use it for factual and methodological errors rather than personal health questions — nothing here can be advice about your situation, and the medical disclaimer explains why.