There is a phrase that shows up constantly in health reels, diet posts, and political talking points. It sounds authoritative, it sounds like evidence, and it is almost designed to resist being checked.
"A study says…"
I want to explain why, whenever I see that phrase without anything else attached, my first reaction is not curiosity — it is suspicion. And then I want to show you what to do about it in two minutes.
The rhetorical move
Citing a study does something clever: it borrows the credibility of science without actually handing over the receipt. Science is trustworthy in aggregate, over time, because it is self-correcting. Individual studies, on their own, are not verdicts — they are opening bets. The whole apparatus of peer review, replication, and meta-analysis exists because single studies are wrong surprisingly often.
When someone says "a study says X," they invoke the institution without giving you any way to interrogate it. You cannot check the journal. You cannot check the sample size. You cannot check whether the researchers declared conflicts of interest. You cannot check whether the finding was replicated or retracted. You are being handed a conclusion and asked to accept it on faith — which is the opposite of how the underlying science is supposed to work.
This is not always malicious. Many people share "a study says" content because they found it compelling and did not stop to think about what was missing. But the rhetorical effect is the same either way: authority without accountability.
What a real citation needs
A citation that lets you actually verify a claim has at minimum five things:
- Authors. Who ran this study? Are they affiliated with the institution making the claim, or with a manufacturer who benefits from the result?
- Publication. Where was it published? A peer-reviewed journal, a conference proceedings, a preprint server, a blog post? These are not equivalent.
- Sample size and population. Twenty college students at one university for three weeks is not the same as a randomised controlled trial with thousands of participants across multiple countries. Both are "a study."
- Date. Scientific consensus moves. A study from fifteen years ago on a fast-moving topic may have been superseded a dozen times since.
- Peer-review status. Was it reviewed before publication, or is it a preprint — a draft shared before independent experts have checked it?
None of this is pedantic gatekeeping. Every one of those details changes what the study actually proves.
Common abuses
Cherry-picking. If I run ten studies and nine find no effect, but one finds the effect I want, I can truthfully say "a study found X" — while hiding the other nine. This is why scientific consensus matters more than any single result.
Preprints presented as settled science. Preprints are drafts. They are useful for rapid sharing, and most of them are never cited after posting because peer review finds problems. "Researchers at [institution] found…" sounds definitive. It may be describing something that was posted yesterday and has not been touched by anyone else in the field.
Misread abstracts. Abstracts are summaries written for other researchers who know the field. Read out of context, they often sound more definitive than the full paper supports. "The results suggest a potential association under certain conditions" becomes "scientists CONFIRM…" somewhere along the way.
A single study versus scientific consensus. This is the most important distinction. Scientific consensus is the accumulated weight of many independent studies, systematic reviews, and meta-analyses, evaluated over time by researchers with different incentives. A single study — even a good one — is not that. The gap between "one study found" and "the evidence consistently shows" is enormous, and it gets flattened constantly in viral content.
How to trace "a study" back to its actual source in two minutes
When you encounter an unsourced "a study says," here is the shortest path to the real thing:
- Copy the claim verbatim and search it in Google Scholar or PubMed (for health claims) or Google with
site:scholar.google.com. If a real paper exists, it usually surfaces in the first page of results.
- Look for a news story about the study. Academic papers almost always get covered when they go viral. The news story usually names the journal and the lead author, which gives you what you need to find the original.
- Check the journal. Once you have a journal name, a two-second search tells you whether it is indexed (Scopus, Web of Science), its impact factor, and whether it has a history of retractions. Predatory journals exist specifically to publish low-quality research for a fee, and they look convincing on a screenshot.
- Search the claim in Google Fact Check Tools (
toolbox.google.com/factcheck/explorer). If it has circulated enough to be checked by AFP, Snopes, Reuters, or any other member of the IFCN, the work may already be done.
The whole sequence takes two minutes if the claim is verifiable. If you cannot find the study after a genuine effort, that absence is itself information.
Why verifAI insists on linked sources — and says "unverified" when it can't find one
When verifAI checks a claim, it runs two things in parallel: Google Fact Check Tools and a live web search. Both are required to produce actual URLs — not summaries, not confident-sounding explanations, but links to the real sources that support the verdict.
This is not a technical nicety. It is the whole point. If I show you a "false" verdict with no source, you have no way to check whether I am right. You are in exactly the same position as someone who reads "a study says" and is expected to just believe it.
When neither path turns up relevant evidence, verifAI returns unverified. The explanation says there is no current public evidence to confirm or deny the claim. Some people find this anticlimactic. I find it honest. The alternative — synthesising a confident verdict without sources — is the same problem I just spent several hundred words describing.
A tool that fights "a study says" misinformation by guessing its verdicts is not a fact-checker. It is just a different kind of misinformation engine.
You can read more about what "unverified" actually means and the cases where it applies in What "unverified" really means. And if you want to see how a real citation chain gets traced from a reel to its source, the case study on "no market without the state" walks through it concretely.
If you want to try the full pipeline yourself, paste a claim or a reel URL into verifAInow.es. If you think a verdict is wrong, I want to know — my inbox is fernandoruedaoliva@gmail.com.