The product page for kēpos™ for Digestion, an $89.99-a-month powder, makes its pitch in two sentences:
As we age, the gut microbiome ages, as well, losing its diversity, resilience, and vibrancy. kēpos replenishes it directly, using the same bioactives human milk uses to build our microbiomes from scratch.
On the company's "Proof" page, the same argument arrives at its payoff: feeding Bifidobacteria delivers "stronger immunity, better food tolerance, a younger gut microbiome."
This is an unusually well-built claim. The ingredients are real molecules — human milk oligosaccharides (HMOs) and human lactoferrin — with real randomised trials behind them. The mechanism copy is broadly accurate. And crucially, the Proof page cites its sources by PubMed ID, which is an invitation to check. We checked. The chemistry holds up; the arithmetic does not; and the ageing claim on top of both was never tested by anything the company cites.
The part that is true
Human milk oligosaccharides really do feed Bifidobacterium in adult guts, and this has been shown against placebo more than once. In Elison et al. (2016), 100 healthy adults were randomised double-blind across ten arms for two weeks: "In total, 77 % of the participants responded to the HMO interventions," a responder being someone whose Bifidobacterium sequence abundance rose by more than 10%. In the placebo group, "none of the eighteen genera changed." Iribarren et al. (2020) reproduced the effect in IBS patients at 10 g/day, and the Stanford RAMP trial reproduced it again in older adults at 5 g/day of 2'-FL.
The microbiome also genuinely changes with age. Wilmanski et al. (2021) looked at more than 9,000 people across three cohorts and found real, measurable age-related shifts.
So the biology is not the problem. Everything that goes wrong here happens in the step from "these molecules feed a bacterium" to "this powder makes your microbiome younger."
The footnotes do not check out
The Proof page runs a stat block — "+50% Average digestion improvement¹ / +77% Reported less bloat¹ / +82% Grew Bifidobacteria levels²" — footnoted to PMID 33512807 and PMID 27719686. Those are the only two citations on the entire site.
PMID 27719686 is Elison. Elison reports 77% of participants responding, not 82%. The 82% does not appear in the paper. Meanwhile the 77% figure on the page is attached to the other footnote, the IBS trial. The two numbers appear to have been shuffled between papers.
PMID 33512807 is Palsson et al. (2020). Elsewhere the site describes "245 completed participants." In the paper, 317 were in the intention-to-treat analysis and 273 completed all assessments; 245 is the number who reported full adherence. And the "+50% average digestion improvement," "77% reported less bloat," "88%" and "59% reduction in bloating" figures are not in that paper's reported results at all — it reports group means for those outcomes, not responder percentages.
A footnoted PMID reads like a settled question. Here, two of them, and every number sitting on top of them is either wrong or untraceable.
We also dropped several figures the site leads with — "1,000+ clinical trials on human milk bioactives," "700+ doctors," "1M+ doses," "4.7 from 480+ reviews." Not because we think they are false, but because none is sourced anywhere and we could not trace any of them. An unsourced round number is not evidence, even when it is probably true.
The company cites the uncontrolled trial and not the controlled one
This is the part that matters most.
Palsson (2020), the footnoted IBS study, describes itself in its own methods as "a prospective, open-label, single-arm clinical trial." No control group, no blinding: "The intervention was open-label, so all patients were aware they were taking the active HMO mix." Some 88.3% of participants also stayed on their IBS medication throughout. The results are striking — IBS-SSS fell from 323 to 144 — and the authors are entirely honest about what that means. Their own conclusion: "These promising results suggest that this novel approach to IBS should be confirmed in a randomized, placebo-controlled trial."
That trial exists. Iribarren et al. (2020) tested the same 4:1 2'FL/LNnT compound, in the same patient group, funded by the same ingredient manufacturer, published the same year — but randomised, double-blind and placebo-controlled, n=61. On symptoms: "No differences in overall GI symptom severity (GSRS-IBS total score) between groups were identified at week 4 or week 8." No between-group difference on IBS-SSS either. And the within-group detail is worth reading twice: "Within-groups comparisons detected a decrease in the severity of bloating and diarrhea in the placebo group at week 4, but no differences in the other groups."
That is not a demonstration that HMOs do nothing for IBS. Iribarren was designed as a dose-finding study — which dose raises bifidobacteria without aggravating symptoms — not an efficacy trial, and its authors do not claim otherwise. But the seller footnotes the trial with no placebo arm and does not mention the one with a placebo arm.
Iribarren also measured what happens after you stop. The 10 g arm's bifidobacteria rose significantly at week 4; by week 8, four weeks after stopping, "no differences in fecal bifidobacteria abundance between groups were detected." The RAMP authors reached the same place independently: "Reversibility of microbiome changes following cessation of supplementation suggests that continuous prebiotic intake might be necessary to maintain benefits." For a $572-a-year subscription, that is a material fact about the product.
Nothing measured a younger microbiome
Only one cited trial was conducted in the population the claim is about. RAMP (2025, Sonnenburg lab) randomised 89 healthy older adults, mean age 67.3, to placebo, 1 g or 5 g of 2'-FL for six weeks. Its primary endpoint was not met: "We did not observe significant changes in CRS between treatment groups." Bifidobacterium did rise in the 5 g arm. But the post-hoc responder analysis points somewhere awkward for the marketing: "93.3% of responders have Bifidobacterium in their microbiome at baseline, compared to only 58.6% of nonresponders." The supplement feeds what is already there. The paper's own framing is hedged — "showcasing the potential of this prebiotic to provide diverse benefits for healthy aging" — and the trial measured no marker of gut ageing in either direction. Neither did Elison (participants aged 19–57), nor Palsson, nor Iribarren.
And the best available description of what ageing actually does to the microbiome points the opposite way from the sales copy. Wilmanski found that from mid-life onward the microbiome becomes more individually distinctive, with "a decline in the core genus Bacteroides" emerging "as a major characteristic of healthy aging." In the 85-and-over group, higher Bacteroides and lower uniqueness predicted worse survival. On that evidence, healthy ageing is not a slide away from a youthful template that a supplement should push back toward. It is observational work, so it licenses no intervention claim in either direction — but it does not support the frame the claim depends on.
The second active ingredient, human lactoferrin (effera®), has thinner evidence still. The one relevant trial is a 28-day exploratory microbiome analysis nested inside a safety and immunogenicity study: 66 adults aged 18–45, 14–15 per arm, funded by the ingredient maker. Alpha-diversity "remained stable across all groups" — no change in richness, evenness or phylogenetic diversity — and the paper does not report Bifidobacterium abundance at all. The authors say so themselves: "clinical studies directly evaluating LF's impact on the adult gut microbiome remain scarce," and "adult-directionality cannot be assumed to translate across life stages." That does not refute the diversity claim. It means the diversity claim has not been assessed at the level it needs.
One practical note on dose. In Elison, the 20 g/day 2'FL arm reported significant increases in nausea, rumbling, bloating, gas and loose stools versus placebo. All adverse events were mild, and safety is the least contested thing about these molecules. But kēpos publishes no dose anywhere on its page, so a buyer cannot tell where on that curve the product sits.
Why D and not E, or C
Not E, because nothing here is contradicted by the evidence in the way E requires. HMOs feed Bifidobacterium; that is demonstrated against placebo three times over. The microbiome does change with age. The mechanism descriptions are accurate.
Not C, because C is for evidence that is genuinely split. On the actual claim — that this powder replenishes an aged microbiome — the evidence is not split. It is absent on one side and negative on the other: no cited trial measured gut ageing, the only placebo-controlled symptom trial of this compound found nothing, and the only trial in older adults missed its primary endpoint.
That leaves D. Real molecules, real trials, honest authors — and a claim assembled on top of them that none of them tested, sold with footnotes whose numbers do not match the papers they point to. The phrase "a younger gut microbiome" appears on the seller's page. It appears in none of the research.