Editorial standards & corrections
Health information should be useful, traceable and clear about its limits. Here are the standards we are applying to Frida’s learning library—and where that work stands.
Updated · Published by ColdAI LLC
Ownership and authorship
ColdAI LLC owns and publishes Frida. We also own the Frida app, so we have a commercial interest in its use. Educational material must not present that interest as independent evidence that the product improves clinical outcomes.
“Frida Editorial Team” is an organizational byline. It does not identify a clinician or imply medical credentials. Individual author and reviewer names, qualifications and affiliations must be verified before publication. We will not invent credentials, advisory relationships or personal experience.
AI assistance, where used, does not count as clinical review. Generated text, summaries and citations require checking. We are still documenting the creation and review history of the existing library; an article’s byline alone should not be read as a claim about whether AI contributed to it.
What a clinical review label means
A clinical review label may appear only when a verified, named professional with relevant qualifications has approved the exact published version. The record must identify the reviewer, review date and scope. An article’s publication date is not a review date.
Clinical claims, medication information and crisis guidance need appropriate human review. Automated checks can help find problems, but they cannot approve clinical accuracy. Substantive changes require renewed review; an earlier approval does not automatically cover a rewritten article.
Where no qualifying review is recorded, the article must say so. Our articles provide general education and do not diagnose a person, prescribe treatment or replace a care team.
Sources must support the actual claim
Our standard is to use current clinical guidelines, regulatory information, systematic reviews and relevant primary research. A citation must lead to the intended source, with its title, authors, date and identifier checked. A working link to a research website is not sufficient.
- Match evidence to the statement. Check the population, intervention, comparison, outcome and limitations. A source about one drug formulation or one group of people cannot automatically support another.
- Explain uncertainty. Separate association from causation, preliminary research from established care, and performance in a research dataset from validation in clinical practice.
- Keep product evidence separate. Research about an AI model, wearable or therapy does not establish that Frida diagnoses, predicts, prevents or treats schizophrenia.
- Identify the jurisdiction. FDA labels describe US regulatory information. Approval, availability and prescribing requirements can differ elsewhere. US information must not be presented as a global rule.
- Verify practical details. Local services, crisis contacts and eligibility need checks against current official sources and a recorded check date.
Examples of sources readers can consult directly include NIMH’s schizophrenia information, NICE’s guideline on psychosis and schizophrenia in adults, and the WHO guidance on ethics and governance of AI for health. Their inclusion here does not imply endorsement of Frida.
Stories and lived experience
Some Frida articles use illustrative composite stories. These do not depict a verified individual and must not be presented as patient testimonials, interviews or evidence of treatment success. The composite label should be visible on the article and wherever its story is introduced.
Genuine personal contributions require documented consent and provenance, with a clear explanation of pseudonyms or changed details. A person’s account deserves respect, but it does not establish what will happen to another person. Medical statements and advice attached to a story still require evidence and review.
Useful pages, maintained carefully
New articles and changes to article titles, summaries, body text or FAQs require a recorded clinical review before the publishing system accepts them. Existing material remains available during the audit, except for the pages placed on hold. Keeping an existing page available is not clinical approval.
A new article needs a distinct reader purpose and useful, verified substance. We do not treat keyword variations, city names or translation opportunities as sufficient reasons to create pages in bulk.
Closely overlapping articles should be assessed for consolidation. Local guides need verified local information; translations need competent language and clinical review for their intended audience. Publication must stay within our ability to check and maintain the content.
Changes to sources, drug labels, service details or clinical guidance should trigger reassessment of affected articles. Material corrections belong in a visible record. Dates must reflect real publication, updates or completed review—not a cosmetic refresh.
Corrections and temporary holds
The following record covers issues addressed on . A correction is not a certification that every statement in an article has completed clinical review. Pages on hold remain unavailable as medical guidance while their review is pending.
-
Reference corrected · 25 September 2026
Schizophrenia AI research guide
The digital-phenotyping reference linked to an unrelated paper. Both the reference and the surrounding description have been corrected to identify the 2016 perspective by Onnela and Rauch. This source is a perspective, not a trial proving clinical benefit.
-
Corrections prepared · Article held for clinical review
Erzofri medication article
Incorrect dosing information has been removed, and misleading initiation language and research citations have been corrected in the article under review. The article remains on hold pending qualified clinical review. The original wording should not be used for medication decisions.
The review uses the US Erzofri prescribing information. The corrected PRIDE study reference concerns a different paliperidone formulation and must not be presented as direct evidence for Erzofri.
-
Temporary hold · Clinical and source review pending
Two command-hallucination articles
We have held these articles because coping suggestions could be misinterpreted in a situation involving harmful commands. The review also covers citation errors and overlapping content. A temporary hold does not mean that people who hear voices are inherently dangerous.
Command auditory hallucinations: what they are, how to cope
Command hallucinations: voices that tell you what to do -
Temporary hold · Clinical review pending
Ketamine and psychosis article
The article’s contraindication framing did not adequately distinguish a formal contraindication from a warning requiring clinical assessment. It is on hold while those distinctions and the wider treatment discussion are reviewed against appropriate sources, including the US Spravato label.
Help us correct an error
If you find inaccurate, unclear or outdated information, send us the article link, the passage in question and a supporting source if you have one. Please avoid including private medical records or identifying details about someone else.
Our support channel is not an emergency service. If you or someone else may be in immediate danger, contact local emergency services.