We checked all fourteen references on a supplement sales page

There is a particular design pattern you’ll recognise once you’ve seen it. A supplement sales page makes a series of confident claims — about inflammation, immunity, sinuses, sleep — and then, somewhere below the order buttons, a heading reading Scientific References followed by a numbered list.

The list does real work. It signals that somebody has done the reading, that the claims above are downstream of evidence rather than copywriting. Most people scroll past it. That is more or less the point: its function is to exist, not to be read.

So we read one.

We picked the reference list on the ProDentim sales page, because we were reviewing that product anyway and because it is among the most heavily advertised oral probiotics sold online. We are an affiliate for it, which we’ll come back to, because it is relevant to how you should read this.

Fourteen references. Here is what they were.

What the fourteen actually were

Five were papers in peer-reviewed journals. These are real. Several are relevant. The research on oral probiotics is a legitimate field and some of these citations point at genuine work within it.

Nine were not. Specifically:

• A news feature in Nature — journalism about science, published by a scientific journal, which is a different thing from a paper
• Two articles in trade press
• Six consumer-facing web posts: Healthline, LIVESTRONG, a dentist’s own blog, a lifestyle magazine, a supplement retailer’s website, and a self-published Medium article

None of those nine are secret or disreputable in themselves. Healthline is a normal thing to read. The problem is the framing: a list headed “Scientific References” that is nine-fourteenths not scientific references is doing something other than citing evidence.

The two that don’t hold up

Two entries fail in a way that’s worth naming specifically, because both are checkable in about two minutes and both were checkable by whoever assembled the list.

A citation whose own details don’t match

One entry is a randomised trial on Bifidobacterium in chronic periodontitis, cited as 2012;45(10):1198-1210.

The study is real. Those volume and page numbers, though, belong to a paper published in 2018 — not 2012. The reference points at genuine research and describes it wrongly.

Six years is not a rounding error. It is the difference between a citation somebody checked and a citation somebody pasted.

A reference to a different ingredient

The second is more consequential. The product contains tricalcium phosphate. The supporting reference beside that ingredient is an article about dicalcium phosphate.

Tri and di. Related compounds, different substances, different properties. The reference does not support the claim it is placed next to — it supports a claim about something else.

What this does and doesn’t prove

It does not prove the product doesn’t work. That’s a separate question, and the honest answer to it is more complicated. Some of the strains in this formula have real published research behind them.

What it shows is narrower and, we’d argue, more useful: the reference list was assembled to look authoritative rather than to be checked. Nobody who verified these entries would have left a six-year discrepancy in a volume number or a citation about the wrong compound.

And that matters for how you read everything above it on the page. The same list is used to support claims about sinuses, allergies, sleep and digestion — claims that go well beyond anything the underlying research addresses. If the evidence layer wasn’t checked, the confidence layer built on top of it hasn’t earned its confidence.

How to run this check yourself

The whole audit took about ninety minutes and required no institutional access. Here is the method, which works on any supplement page with a reference list.

1. Count the ratio first Before reading anything, sort the list into peer-reviewed papers and everything else. Blogs, press releases, retailer pages, news articles and Medium posts are not evidence. If the non-paper share is over half, you have learned something before reading a word.
2. Check that each reference matches its own metadata Search the paper title. Confirm the journal, year, volume and page numbers agree with what’s cited. Mismatches are common and they are diagnostic — they tell you the list was copied rather than compiled.
3. Check the reference actually supports the adjacent claim This is the step almost nobody takes, and it’s where the real failures live. A paper about a related compound, a different strain, a different population or a different dose is not support. Read at least the abstract.
4. Check the strain, not the species This one is specific to probiotics and it matters enormously. Probiotic effects are strain-specific. Evidence for Lactobacillus reuteri DSM 17938 is not evidence for a different L. reuteri strain. A product that names only the species is not making a claim the research can be matched to.
5. Check the dose is disclosed Research on an ingredient at a studied dose says nothing about a product containing an undisclosed amount of it. A combined CFU total across three strains, with no split, cannot be compared to any trial.
6. Distinguish “studied” from “proven” Most probiotic trials are small. Effects are usually modest. A page that cites a small trial as though it settles the matter is overstating it — and so is a page that dismisses the field entirely.