Skip to content
Submit incident
AII-1001-95679
Well documentedThree or more independent publishers report the core facts.

Federal 'Make America Healthy Again' Report Contains Fictitious and Erroneous Citations

A federal health report released by the Trump administration contained dozens of erroneous citations, including at least seven entirely fictitious studies. Evidence including "oaicite" URL markers suggests ChatGPT was used to generate citations, a common source of AI hallucinations.

22 May 2025 Β· White House Β· Reliability and hallucination
22 May 2025
AII-1001-956793 sources ↓
Quick facts
Category
Reliability and hallucination
Country
United States
Source
nytimes.com
Date
22 May 2025
Drafted by
Raidu Incident Detection

What happened

The Make America Healthy Again Commission report, released by the Trump administration in May 2025, contained dozens of erroneous citations including broken links, wrong issue numbers, missing or incorrect authors, and at least seven entirely fictitious studies. Investigations by NOTUS and The Washington Post identified false citations on topics including direct-to-consumer drug advertising, mental illness, and asthma medications in children.

Several URLs in the citations included 'oaicite,' a marker that OpenAI applies to ChatGPT responses, suggesting the AI tool was used in developing the report. The Washington Post found that at least 37 of the 522 citations appeared multiple times throughout the report.

Reported impact

Katherine Keyes, a Columbia University epidemiology professor, was listed as the author of a paper on mental health and substance use among adolescents that she had not written. The errors raised concerns about the rigor of the report and its lack of proper citation verification before release.

Index analysis

Experts noted the citation errors were characteristic of generative AI hallucinations, which have produced similar issues in legal filings and scientific literature. The false references indicate insufficient review of the report and its bibliography before publication.

Governance controls

By midafternoon on Thursday, the White House uploaded a corrected version of the report with some oaicite markers removed and some nonexistent sources replaced with alternative citations. The Department of Health and Human Services characterized the errors as 'minor citation and formatting errors' while maintaining that the substance of the report remained unchanged.

Classification

Category
Reliability and hallucination
Country
United States
Harm status
Potential harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Public interest, Reputational
Harmed
Government, General public
Failure modes
Hallucination, Fabricated citation
Consequences
Content removal
AI systems
ChatGPT

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

nytimes.com
Federal 'Make America Healthy Again' Report Contains Fictitious and Erroneous Citations
theverge.com
Federal 'Make America Healthy Again' Report Contains Fictitious and Erroneous Citations
washingtonpost.com
Federal 'Make America Healthy Again' Report Contains Fictitious and Erroneous Citations