Method

What "checked" means on this page

A citation passes our check when two things are true. First, its identifier resolves: a PMID lands on a record at PubMed, run by the National Library of Medicine, or a DOI resolves at doi.org and exists in CrossRef. Second, the record it lands on matches the sentence it supports: the named authors, the journal, the year, and the finding all line up with what the marketing page says the study shows. A citation fails when the identifier leads to an unrelated paper, when the named authors do not match the real paper's authors, when the identifier returns an error, or when a link is assembled wrongly, such as a PMC article number placed in a PubMed URL slot.

Two things do not count as checkable. A link to a journal's homepage is not a citation to a study, because there is no specific paper to verify. And a screenshot of a paper, with no link or identifier, can be verified only by reconstructing the reference by hand, which our Aurien review did and most shoppers will not.

Scope matters, and each row below states it. Some brands received a full audit of every checkable identifier on the page, such as Kreain, where all 18 were run. Some received a spot-check, such as Nathan & Sons, where 8 of roughly 39 were run and the page says so plainly. Brands that cite nothing receive neither, because there is nothing to run; for those, the documented finding is the absence itself, and our pages are explicit that citing nothing is a materially different, milder failure than printing identifiers that resolve wrong. Every finding in this reference is quoted from the dated review page linked in its row, checked on the date that row states. We added no new checks for this page except one: our own, which was run in August 2026 across the entire site and is documented in its own section below.

The table

Every brand, including us

Every quoted phrase in this table comes verbatim from the linked review page, as checked on the date given in its row. The category also has a companion page grading what each brand discloses, beyond citations: the disclosure grades.

Group one

Citation lists where the identifiers lead elsewhere

Two storefronts in this category print reference sections that look like the bibliography of a thesis: numbered lists, author names, journal titles, years, identifiers. Both were audited by our review pages against PubMed and CrossRef, identifier by identifier, and both pages publish the working notes so anyone can re-run the checks.

Nathan & Sons / Underbrush. Our Nathan & Sons page documents that "the product page carries numbered reference sections, one per ingredient, with roughly 39 linked citations." On August 27, 2026, eight of those were run against PubMed and CrossRef. Three resolved cleanly: Najibfard (PMID 22403978), Schlagenhauf (PMID 30701704), and Landeo-Villanueva (DOI 10.3390/antibiotics12020369). Five did not. The page's summary line is seven words long: "5 of 8 spot-checked citations misresolve."

The five failures, exactly as our audit table records them. A citation reading "Makinen KK, et al. Xylitol for preventing dental caries. Cochrane Database Syst Rev, 2015" carries PMID 25867215, which resolves to "an unrelated paper on peri-implantitis by Shi and colleagues": the page's verdict column reads "Wrong record." A mastic-gum citation names "Vukovic AM, et al." in Archives of Oral Biology, 2006, but "The 2006 Arch Oral Biol mastic-gum paper is by Aksoy, Duran and Koksal": "Wrong authors." A calcium carbonate citation names "Zero DT, et al." in Caries Research, 2003, but "The matching Caries Research 2003 paper is by Cury and colleagues": "Wrong authors" again. A mint-oil pilot study is real, PMID 39203382 by Abdelrahman and colleagues, "but attributed to different, unnamed authors on the page": the note reads "Real study, wrong attribution." And the gum-arabic citation is "A PMC article ID placed in a PubMed URL slot; the link redirects to an unrelated record": "Malformed link." Our page is explicit about scope: roughly 31 of the 39 citations were not checked, so this is a spot-check, not a full audit. One housekeeping note: this forensic audit lives on the Nathan & Sons page; our separate Underbrush comparison page lists only Minvelle's own sources and does not repeat it.

Kreain. Our Kreain review went further, because Kreain's page allowed it: "Of the 39, 18 carry a checkable identifier... On August 28, 2026, we ran all 18 against PubMed and CrossRef, one by one. Ten checked out. Eight did not." Three of the eight failures are the same ones documented on the Nathan & Sons page: the Makinen citation with PMID 25867215 resolving to "Shi et al., a plasma-treatment study on peri-implantitis... Unrelated to xylitol; wrong author; not the Cochrane review," the Vukovic mastic-gum citation where "The real Archives of Oral Biology 2006 paper is by Aksoy, Duran, and Koksal. The named author does not match," and the Zero DT calcium carbonate citation where "The real Caries Research 2003 paper is by Cury et al. The named author does not match."

Five further failures are Kreain's alone, and the review documents where each identifier actually leads. A "Zhang S" bentonite citation with DOI 10.1016/j.clay.2018.01.025 resolves to "A glycerol-carbonate catalysis paper. Not about bentonite's antibacterial properties." A "Pochapski J" citation resolves to "A bone-wax cement paper. Not the study described." A "Müller M" citation for bentonite binding toxins resolves to "A cyanotoxin analytical-chemistry paper... Not about bentonite." A "Rahman I" zinc mouthwash citation resolves to "A tooth-loss prediction modeling paper. Not about zinc mouthwash." And a "Kumar P" citation resolves to "Nothing. The DOI returns an error at doi.org and does not exist in CrossRef." The gum-arabic citation carries the same "malformed URL putting a PMC number on the PubMed domain" as Underbrush's. What that overlap means is a question the next section takes up, in exactly the words our pages use.

The fingerprint

Two storefronts, the same errors

An error can be shared by coincidence when it is generic. These are not generic. Our Kreain review names the pattern directly: "several of Kreain's errors are not merely similar to Underbrush's. They are the same errors, distinctive down to the digit." Here is the overlap as our two pages document it, side by side.

On what this overlap means, our Kreain page draws no conclusion about cause, and neither does this one: "we do not assert what explains them." The overlap is documented, the identifiers are printed, and every check can be re-run by anyone with a browser. That is where both pages stop, and it is where this page stops too.

Group two

Real studies, documented caveats

Four brands cite at least one real, correctly reproduced study. That deserves saying plainly, because in this category it is the exception. It also deserves the caveats our pages attach, because a real study can still be doing the wrong job on a marketing page.

Aurien shows three studies on its About page, and our review confirms each: "Real and correctly reproduced. PMID 22403978, confirmed against PubMed" for Najibfard and colleagues, "Real and correctly reproduced. DOI 10.1016/j.jdent.2024.105402, confirmed against CrossRef" for Alajlan and Baysan, and "Real and correctly reproduced. PMID 25506416... confirmed against PubMed" for Pepla and colleagues. The page's explicit verdict: "There are no fabricated citations here. Nothing was invented, no author was swapped onto someone else's paper, no identifier leads somewhere other than where it claims." Two caveats follow. None of the three papers is about chewing gum: two are toothpaste and lesion studies, one is a general dentistry review. And all three appear as unlinked screenshots, which matters because, as our page puts it, "a screenshot is not a citation." There is also one finding that sits apart from the About page: on the advertorial at the brand's /pages/adv1, a remineralization claim is attributed to a "Harvard Health Study," and our page's verdict is blunt: "'Harvard Health Study' is not the name of a study; it is not the name of anything." No author, year, journal, link or identifier accompanies it.

Milliways cites real science for its plastics angle, and our review credits it before flagging how it is used: "The number is real, the study is real, and the paraphrase quietly strips the conditions off it." The study measured particles accumulating in saliva across an hour of chewing; "the marketing turns that into a per-piece leaching figure." Our page also documents an omission: "Milliways quotes the Belfast study without its conditions and does not mention the UCLA pilot at all," referring to a 2025 UCLA pilot study that complicates the natural-versus-synthetic gum-base picture. For the dental claim, the citation practice is simpler: "The homepage says the gum 'keeps teeth healthy, prevents tooth decay', credited to xylitol, yet no xylitol dose per piece is published and no study is cited on the site."

Xylichew is the category's strangest citation story, and it is entirely real. Our review: "the FAQ does cite references, three of them: a JECFA safety confirmation from 1983, an FDA confirmation from 1986, and one, exactly one, peer-reviewed journal article. That article is from the Journal of the American Veterinary Medical Association, 2006, and it is about xylitol toxicity in dogs." Nothing misresolves; the citation is correctly described. The problem is placement: "On a page claiming benefits for human ears, human bones and human candida, the only journal citation is a paper about poisoned dogs." The beyond-teeth claims themselves carry, per our page, "No citation... attached to any of it."

Dentagum cites real literature, including a 2024 in vitro study and a review in Evidence-Based Dentistry, and our review documents no misattribution at all. The finding is scope: "The eye-catching research that gets cited for eggshell remineralization is almost always about eggshell-derived hydroxyapatite, the lab-made version," while a pinch of raw eggshell powder is a different proposition. And: "most of the evidence is preclinical. The studies are overwhelmingly in vitro or in animals... Controlled human trials showing that an eggshell gum rebuilds enamel or prevents cavities over months of real use are not what these papers provide." Real papers, honestly cited, describing something other than the product as marketed.

Group three

Claims with nothing behind them

Eleven brands make claims and cite nothing checkable for them. Our pages are consistent about what this is and is not: as the Chewsy and Gud Gum reviews both put it, "Silence is a disclosure gap; a broken citation is a misrepresentation." Nobody on these pages is being sent to a study that says something other than what they were told, because nobody is being sent anywhere. That is a real failure, documented below brand by brand, and it is a milder one than the previous section's.

Larineco is the sharpest version, because the claim is so specific. Our review: "The 'Backed by thousands of peer-reviewed studies' line has no citation, footnote or link anywhere on the site... A direct query on PubMed... for the term 'larineco' returns no results." The company's own Terms of Service state that "Any research references relate to individual ingredients and may not apply to our specific product formulations," which our page treats as the company's own admission that nothing backs the storefront claim. Our page is careful about the category: this is a documented cites-no-studies finding, not a misresolving-citation finding, because there is nothing to run.

Chewsy was searched thoroughly. Our review: "Chewsy's pages carry no citations at all. We searched the full source of the homepage and the product pages for PMID, for doi.org, for the words study, clinical, reference and footnote, and for any superscript numbering. Nothing." The same review draws the category line explicitly: "Nothing is fabricated. There are no citations on these pages, and that is a real gap, but it is emphatically not the same failure as a page that lists PMIDs resolving to unrelated papers... Chewsy makes claims and offers no support."

Halo is a distinct sub-case: no PMIDs, but numbers that sound like they came from somewhere. Our review: "The science page displays a 48% and a 99% bacteria-reduction figure with no institution, no sample size, no date, no publication and no link, so nobody outside the company can verify either number." The figures are "footnoted to an unnamed human trial and in vitro saliva tests." Our page explicitly declines to go further: "We do not claim the numbers are invented; we state exactly what is missing." Since no identifiers are printed, there is nothing to resolve against PubMed; the gap is unverifiable in-house statistics, not misattributed literature.

Gud Gum cites nothing anywhere, and the gap is deepest exactly where the claim is biggest. Our review: "There are no fabricated citations, because there are no citations... Its dental page states a mechanism and stops." For the hydroxyapatite line specifically: "There is no study cited, not for this product, not for this dose, not for hydroxyapatite in gum at all," and that product has, "alone among the store's products, no ingredient list."

The remaining seven, in the words of their review pages. PUR: "No study named or linked anywhere we found on the site," a "Verified gap; xylitol's real evidence exists, PUR just doesn't cite it." Spry: "'Clinically proven, unique, dental health benefits' and 'numerous clinical studies have shown' appear on the product page with no study named, no journal linked, no number attached," and "a page that gestures at 'numerous studies' without naming one is asking for trust it could have replaced with a link." Vanman: "no clinical evidence of any kind cited," and "Strong verbs demand strong receipts, and no receipts of any kind are cited: no finished-product study, no ingredient-level citation, nothing." Weldental: "No clinical study cited anywhere, finished-product or ingredient-level," with no particle size and no milligram dose published either. Crait: "'Scientific studies show hydroxyapatite helps strengthen enamel, reduce sensitivity, and support remineralization', with no study named or linked anywhere on the site," plus a certification claim that "belongs to the supplier's ingredient family, not to Crait's finished gum." Trident White: no study cited by the brand for its whitening claim, and the third-party study everyone reaches for, per our page, "concerns an ingredient this gum, as currently listed, does not declare." CaviChew, credited by our page for candor: its listing "openly borrows the category's evidence rather than inventing its own," answering a results question with "Studies on other remineralizing gums have shown remineralization support with similar use."

Two discontinued brands close the set, and both close it well. Glee Gum: "We found no dental, remineralization or clinical-efficacy claims anywhere on gleegum.com, no dentist endorsements, no cited studies, no white-coat photography, no dose disclosure of any active ingredient, because there was no active ingredient being marketed." Our page frames the result as integrity: "There is nothing to check... Glee Gum solved that problem in the bluntest possible way." Project 7 "was a flavor brand, not a dental brand... Nobody chewed Fairytale Fruit because of a remineralization study, and the brand, to its credit, never claimed anyone should." A brand that makes no claim owes no citation, and our pages record both of these as points in the brands' favor. One more scope note for completeness: ApaCare appears in our coverage as a comparison page, not a forensic review, and that page runs no citation audit. Its documented finding is about the label, quoted from the brand's own product page: the printed ingredient list has "no hydroxyapatite named in it and no per-piece dose," while the product description names hydroxyapatite. As our page says of its own care here: "Both of those are facts about what is published, not a claim about what is in the tin, and we would update this paragraph the same day ApaCare published a dose or an amended list."

Do it yourself

How to check a citation in five steps

None of the checks on this page needed special access. Every one runs on a phone, in a minute or two, on public databases. Here is the method, in the order we use it.

1. Find the identifier. Look for a PMID (a string of digits, PubMed's index number) or a DOI (starts with "10.", the publisher's permanent identifier). A citation with neither, a journal-homepage link, or a screenshot cannot be checked directly; treat the claim it supports as unverified until you reconstruct the reference by hand.

2. Resolve it. For a PMID, go to pubmed.ncbi.nlm.nih.gov and paste the number into the search box. For a DOI, type doi.org/ followed by the identifier into your address bar. If the DOI returns an error page, the citation fails at step two, which is exactly what happened with the "Kumar P" identifier our Kreain review checked.

3. Match the authors and year. Read the record that comes up. Do the authors on the marketing page appear on the real paper? Is the year right? Is the journal the one named? This single step caught the Makinen, Vukovic and Zero DT errors documented above: in each case the paper existed, but the people named on the gum page were not its authors.

4. Match the finding to the sentence. Read at least the abstract and ask one question: does this paper show what the sentence next to the citation says it shows? A real paper about toothpaste does not support a gum claim by itself, a study of particles in saliva over an hour is not a per-piece leaching figure, and a veterinary toxicity paper is not evidence about human ears. This is the step where real citations doing the wrong job get caught.

5. Note what you found, with the date. Pages change. If a citation fails, screenshot the page and the database record with the date visible. If you write to the brand, or to us, that pair of screenshots is the entire case, and it is enough.

Questions, answered

The things people ask about this