Incident
Incidents · MAHA report, 2025
The 2025 MAHA report cited studies that do not exist
A government report with hundreds of citations included references to studies that were not published. The failure is bibliographic: the footnote names a paper, and the paper is not there.
What was reported
On 29 May 2025, NOTUS reported that the Make America Healthy Again Commission’s report on children’s health cited more than 500 studies and other sources, and that seven of those cited sources do not appear to exist. One citation listed epidemiologist Katherine Keyes as an author of a paper on adolescent mental health in JAMA Pediatrics. Keyes told NOTUS: “The paper cited is not a real paper that I or my colleagues were involved with.” NOTUS also described two citations, attributed to Shah and to Robert Findling, that do not appear in the journals named in the footnotes.
Reuters reported the same day that it had independently confirmed two of the erroneous citations. White House press secretary Karoline Leavitt described the citation errors as formatting issues. NOTUS also quoted an HHS spokesperson, Emily Hilliard, calling them minor citation and formatting errors. Reuters reported that the government posted a corrected version later that day. A NOTUS follow-up said all seven of the citations it had flagged were replaced in the updated file.
What is not established here
Reuters wrote that the errors raised questions about whether the report had relied on artificial intelligence. That is a question the coverage raised. It is not a finding that a specific model wrote the footnotes. This page does not adopt that explanation.
Later coverage found further problems in the corrected file. Those details are left out here unless you go to the follow-up NOTUS article linked below, which describes replacements that do not necessarily support the report’s claims. This summary stops at the points the three linked reports agree on.
What a tool like RefCheck would be looking at
The cited items were ordinary-looking scholarly references: authors, a journal, a year, in one case a DOI that did not lead to the named paper. That is the input RefCheck is built for. It would search Crossref and OpenAlex, and PubMed or Europe PMC for a biomedical citation, and return likely fabricated when nothing matches, or real with errors when a DOI points at a different work.
RefCheck was not used on the MAHA report. A database miss is also not a legal finding. Chinese-only journals and some books stay unconfirmed because the indexes are incomplete. The Keyes example is the clearer pattern: a named researcher saying the paper is not theirs, and the journal issue not containing that title.
Typical cases
A named author who did not write the paper
Keyes’s statement to NOTUS is the useful test. If the person the footnote credits does not know the paper, stop treating the footnote as a source.
A DOI-shaped link that does not open the article
NOTUS described a nonfunctional DOI on the Keyes citation. A string that looks like a DOI is not a resolved one. Compare the landing page with the title in the footnote.
A corrected PDF the same day
Reuters reported a corrected version later on 29 May, and NOTUS reported that the seven flagged citations were replaced. Replacing a footnote is not the same as showing that the new source supports the sentence.
Sources
- NOTUS, 29 May 2025 — Seven cited sources that do not appear to exist, including the Keyes example, and the White House description of formatting issues.
- Reuters, 29 May 2025 — Reuters confirmed two of the erroneous citations. The White House attributed the errors to formatting issues and said a corrected version was posted the same day.
- NOTUS follow-up — The seven citations NOTUS had flagged were replaced in an updated file.
This page only states what these sources say. It does not say RefCheck was used on the document, and a later correction is not summarized here unless one of these sources describes it.
How it works
Three steps for this kind of reference
01PasteOne reference per line, or a numbered list. Keep the authors, title, year, and the DOI or arXiv id if you have them.
02Cross-checkA scholarly footnote is matched to Crossref, OpenAlex, and, for biomedical items, PubMed or Europe PMC. No match is reported as likely fabricated.
03Read the verdictReal, real with errors, likely fabricated, or unconfirmed. A corrected citation is shown when the record disagrees, and a retraction is flagged when one is on record.
Frequently asked questions
Did the White House say the citations were fake?
No. Leavitt and an HHS spokesperson described them as formatting issues. NOTUS reported that seven cited sources do not appear to exist, and Reuters said it confirmed two of those errors.
Was the report written by ChatGPT?
The articles linked here do not establish that. Reuters said the errors raised the question of artificial intelligence. This page does not answer that question.
Would RefCheck have caught the Keyes citation?
It is the kind of reference RefCheck checks: a journal citation with no matching record, or a DOI that does not belong to that title. RefCheck was not run on the report, so this page does not claim a verdict for that file.
How many citations were wrong in total?
NOTUS reported seven that do not appear to exist, inside a bibliography of more than 500 sources, and also described broken links and misstated conclusions. This page does not add a larger total.
What does a check cost?
3 free reference checks without an account, plus 5 more free after you sign in. Then each reference costs 1 AILatest credit. Failed lookups are refunded. Credits last 1 year and the expiry extends on every top-up. Packs are $1.99 for 200, $4.99 for 600, and $9.99 for 1,500. There is no subscription.
Cite only what's real.
3 free reference checks without an account, plus 5 more free after you sign in. Then 1 credit each.