Accuracy
Corrections
Material corrections to published content are logged here, newest first. How content is checked in the first place is described in How We Review Content. To report an error, use the contact page.
2026-08-14
Study summaries corrected or withdrawn after a faithfulness audit
An internal audit has now compared every one of the 7,960 AI-drafted study summaries on this site against the abstract of the study it describes, with two independent model reviewers on each. It raised 366 findings, and all 366 summaries have been rewritten so that every statement is traceable to its abstract. The clearest group described what hyperbaric oxygen therapy does for studies whose abstracts never mention hyperbaric oxygen at all; those summaries now say so plainly. Others overstated a hedged finding, gave a figure or a treatment window the abstract does not contain, or reported a source's opinion as if it were a result. A smaller number are also withdrawn from the search index and excluded from citation in new articles, where the source is not evidence for the claim being made. Withdrawn pages stay reachable by direct link, so a summary is corrected rather than hidden.
2026-08-13
Severe anaemia shown as an investigational use in the research insights
The research insights pages grouped severe anaemia with investigational conditions. Severe anaemia caused by exceptional blood loss is a recognised indication on the UHMS list, not an investigational use. The category was also labelled in American English. Both are corrected and the published insights data now describes the indication as recognised.
2026-08-13
Per-province facility counts disagreed with the directory
The provincial dataset behind the per-province pages listed 3 private facilities in British Columbia and 15 in Ontario. The verified facility directory holds 2 in British Columbia and 16 in Ontario. The provincial dataset now matches the directory, which is the source of truth for facility counts.
2026-07-19
Four study summaries withdrawn for claims their sources do not support
Four AI-drafted study summaries were withdrawn from the search index after review found statements their source abstracts do not support, including one that gave a 72-hour treatment window appearing nowhere in the cited article. The summaries were de-indexed the same day and excluded from citation in new articles.
2026-07-10
Hearing-loss subgroup threshold corrected to match its source
The sudden sensorineural hearing loss physician referral guide stated that the benefit of adding HBOT was greater in patients with initial hearing loss of 81 dB or more, citing Rhee et al. (2018). The cited meta-analysis defines its severe-to-profound subgroup at 70 dB or more; no 81 dB threshold appears in the article. The page now reads 70 dB or more, matching the source.
2026-07-09
Facility count in llms.txt lagged the directory
The machine-readable llms.txt summary briefly stated 32 verified facilities after a new facility raised the directory to 33. The regeneration process was updated to read the live directory file. The fix reached llms-full.txt, but llms.txt itself still showed the stale total as of the next verification pass and the correction there has not yet been confirmed live.
2026-07-07
Study records relabelled after a study-type audit
An internal audit of study-type labels found a batch of research records labelled as randomised controlled trials that were in fact protocols, reviews, or other designs. The records were reclassified to their correct study types, and study-type counts shown in the insights analyses reflect the corrected labels.
Entries before July 2026 predate this public log. Material corrections to published content are logged here. Routine editorial polish (typos, layout, spelling normalisation, wording that does not change a factual claim) is not logged.