‘When links lead to nowhere, or when a researcher denies having written a work cited by the report, we have a real problem.’
THE word of the day is “oaicite,” a code attribute that means little to mortals, but is a marker automatically generated by OpenAI to announce to the world that information was generated by ChatGPT.
There is nothing wrong with using AI, provided that it does not fully surrender human autonomy and oversight, one of the principles of ethical AI use adopted by the United Nations. In other words, AI makes suggestions very, very quickly, but human users are supposed to confirm manually.
These oaicite markers, says ChatGPT, are not meant to appear in finalized documents. Their presence suggests that AI-generated content was incorporated into the report without proper editing or verification. Oaicite, or OpenAI Citation, is an editor’s, or a teacher’s, assistant that screams of the absence of human supervision.
And scream it did last month when the Washington Post reviewed 500-plus footnotes in the initial version, released May 22, 2025, of the White House’s “Make America Healthy Again” (MAHA) report, titled Make Our Children Healthy Again Assessment. The report was in fulfillment of a requirement by President Donald Trump for the MAHA Commission to, among others, “assess the threat that potential over-utilization of medication, certain food ingredients, certain chemicals, and certain other exposures pose to children with respect to chronic inflammation or other established mechanisms of disease, using rigorous and transparent data, including international comparisons.”
“Some references,” noted the Post, “include ‘oaicite’ attached to URLs — a definitive sign that the research was collected using artificial intelligence.”
ChatGPT gives an example of an oai-cited portion of the report: “A study conducted by researchers at Harvard University suggests that glyphosate exposure correlates with higher rates of chronic illness in children [oaicite:1].” The AI tool says: “This placeholder — [oaicite:1 ]— was meant to link to a web source retrieved by an AI tool like ChatGPT but was not replaced with an actual citation, revealing that the content may have been generated or assisted by AI without proper human review.”
The first MAHA report drew fire for false or fake citations, listing the wrong author and non-existent reports. NOTUS (for News of the United States) noted that the MAHA report cited more than 500 sources but found them to be “rife with errors, from broken links to misstated conclusions. Seven of the cited sources don’t appear to exist at all.”
In an academic setting, this could be the sort of problems college professors might encounter from reckless freshmen, who are admonished, or given a very low grade, even an F, and/or are subjected to disciplinary action. We did not expect this from a White House commission whose members include Cabinet secretaries (among them Robert F. Kennedy Jr.) and heads of agencies like the Food and Drug Administration and the National Institutes of Health.
The White House spokesperson, Karoline Leavitt, a week later dismissed the issue, saying that “formatting issues” in the original MAHA report were being addressed and the report would be updated. The official website now carries only the updated version from which the word oaicite has been thoroughly scrubbed.
Correspondence epistemology, or the correspondence of truth, advances the belief that truth is connected to reality. The idea that you have money in the bank becomes true when it is backed by a balance statement. The fuel gauge on a vehicle’s dashboard reflects the truth if the display accurately reflects how much gasoline is in the tank. News editors check reporters’ claims — one by one.
I know many teachers who check student papers, citation by citation. That is why while most of us enjoy teaching per se, marking papers can be a real, but necessary, pain. We make sure that each “fact” presented in a paper comes with an in-text citation, which must match an entry in the list of references at the end of the paper. Vice versa. We also check if those references indeed exist. Finally, we find out if the reference is applied properly to the claim. Citation by citation. Reference by reference. We are thankful for the Internet and DOIs (digital object identifiers) that allow us to look up a wide range of academic and scientific literature. But this still takes time.
And yes, of course, style is very important. Depending on the discipline or publication, we might require one from an alphabet soup of referencing styles like APA, MLA, Chicago/Turabian, Harvard, Oxford, AGLC or IEEE. I teach in a department of communication, which is still debated whether it is already a discipline or just a field of study. We might prescribe APA in one major program and MLA in another. That, to me, is a formatting issue.
When links lead to nowhere, or when a researcher denies having written a work cited by the report, we have a real problem.
A CBS journalist has objected to Leavitt’s spin. “To say that it is just about formatting errors … is not just sloppiness but really unethical behavior,” said Céline Grounder, who is also an associate professor of medicine at New York University. “Making up a baloney report about America’s health using AI … that is just not ethical. It is not something the US government should be engaging in.”
In its assessment, the MAHA report looks into the “chronic disease crisis” plaguing many young Americans that needs to be addressed.
But for a major report that could influence not only US policy to trigger such a reaction over fundamental issues goes beyond dotting i’s or crossing t’s. It puts whatever significant recommendations under a dark, heavy cloud of doubt.
As Dr. Grounder puts it, “the details matter.” And countless mothers, editors and teachers will agree.
