On May 22 the federal Make America Healthy Again Commission released a 78-page report on childhood chronic disease, attributing rising rates to ultra-processed food, chemical exposures, lifestyle and over-prescribing. Today the news outlet NOTUS reported that seven of the studies it cites do not exist; the Washington Post then found dozens of footnotes carrying a text marker associated with AI-generated references, and at least two named researchers said papers attributed to them were not theirs. The White House called the problems formatting issues and says the report will be updated. We are not going to argue about the report's politics. We are going to show you how to catch a citation like that yourself, because the same thing happens every day in longevity marketing, and nobody issues a correction.
The ninety-second check. First, find the paper, not the claim: search the title and authors in PubMed or Google Scholar. If it has no DOI and no listing, stop; it may not exist. Second, confirm the journal issue and page numbers match a real entry — fabricated references often pair real authors with a plausible title. Third, read the abstract and answer four questions: what species, how many participants, how long, and what was measured. Fourth, check that the paper says what the text citing it says; the most common failure is not a fake citation but a real one that does not support the sentence in front of it. That is the entire method. It works on government reports, supplement websites and podcast show notes equally well.
Ben’s takeI have read pitch decks whose citations dissolve on contact — a real journal, a real-sounding title, no paper behind it. The tell is a reference that is too neat: a perfect title for the claim, no link, no DOI. Whether the source is a founder or a government commission, the fix is the same. Click through. Every time.
Loree’s takeI pull the primary paper for every study a patient brings me, and the gap between what they were told it showed and what it shows is the most reliable finding in my practice. Fabricated citations are the extreme case. The ordinary case is a real study, in mice, for twelve weeks, measuring a biomarker, described as proof. Both fail the same four questions.
Why it matters to a recovering optimizer. The report's broad concern — that ultra-processed food and chemical exposure contribute to chronic disease — has serious evidence behind parts of it. That is a separate question from whether its footnotes are real, and conflating the two is how bad citations survive: people who agree with the conclusion stop checking. The habit we are asking for is to check anyway. Especially when you agree.
Never repeat a health claim you haven't clicked through to the source. If the link doesn't exist, neither does the claim.