For skin, strength, muscle and injury, the honest answer is no. A 2023 trial ran 15 g a day — the dose this site recommends — for 15 weeks, imaged the tendon by MRI, and found no difference from placebo. What survives is narrower: tendon size and stiffness at 15 to 30 g a day, with hard training.
The trial that failed to find anything at that dose is not an outlier a competitor dug up; it is one of the larger, longer and better-imaged trials in the whole literature, and it is the first thing on this page rather than a line near the bottom.
What follows is the state of the evidence, one question at a time, with the sample size and the funding attached to every result. Four of the answers are negative. One is not. The order is deliberate: the negatives come first, because a brand that gives you the positive first has already told you what it wants you to do.
Does collagen actually work?
Not for the four things it is mostly sold for. Independently funded skin trials show no effect. A 2026 systematic review grades the evidence strong against any effect on muscle strength. Thirty grams after training does not raise muscle connective protein synthesis. And no randomised trial has measured injury rates at all. One finding does hold: tendon cross-sectional area and stiffness increase at 15 to 30 g a day, in people who are already training hard.
The whole page is in that paragraph. The rest of it is the working, and the working is where the qualifiers live.
| The question | What the best evidence says |
|---|---|
| Skin, hair, nails | No effect once the funder is not selling collagen |
| Strength | Strong evidence against — GRADE A |
| Muscle | No rise in connective protein synthesis |
| Injury | Never measured in any trial |
| Tendon size and stiffness | Strong evidence for, at 15 to 30 g with loading |
| The same tendon question, by MRI, at 15 g | No difference from placebo |
The last two rows disagree with each other. That is not an editing mistake and it is not resolved further down. It is the actual state of this literature, and any page that reports only one of those two rows has chosen a side rather than reported a field.
What happened when a trial ran 15 g, the dose TARE recommends?
Nothing. Balshaw and colleagues, Med Sci Sports Exerc 2023;55(11):2083–2095, gave 39 men 15 g of collagen peptides or placebo daily through 15 weeks of lower-body resistance training and measured the patellar tendon by MRI rather than ultrasound. No between-group difference was detected on any tendon measure — area, stiffness, modulus, elongation or strain. Both groups improved from the training itself.
Three details make this the hardest single result in the collagen literature to argue with, and each of them is a reason it is usually left out.
- It used our number. Fifteen grams a day, which is the figure this site recommends and the figure the famous 2017 study made famous.
- It ran longer than most of the positive trials. Fifteen weeks of supervised resistance training, three sessions a week.
- It measured the tendon by MRI. Most of the favourable tendon results come from ultrasound. MRI is the harder instrument to argue with, and it is the one that returned nothing.
It is also not an industry trial. The pattern across this field is that the favourable tendon results skew towards industry funding and this null does not, which is a pattern worth holding in mind before the next section rather than after it.
Does collagen do anything for skin, hair or nails?
Myung and Park, Am J Medicine 2025;138(9):1264–1277, pooled 23 randomised trials and 1,474 people, then split them by who paid. Across all 23 the pooled effect was positive. In the trials not funded by a supplement or pharmaceutical company, there was no effect on hydration, elasticity or wrinkles, and the high-quality studies showed no significant effect in any category. TARE makes no skin, hair or nail claim of any kind.
Read the shape of that result rather than the direction of it. The pooled number and the independently funded number are calculated from the same trials. The only thing that changed between them was the question who paid for this, and the effect did not survive the question.
The authors’ own closing sentence is that there is currently no clinical evidence to support the use of collagen supplements to prevent or treat skin ageing. That is the largest market in this category by a wide margin, and it is the one TARE is permanently locked out of by its own claims floor. This paragraph is the whole of what this site will ever say on the subject.
Does collagen make you stronger?
No. The 2026 systematic review in J Funct Morphol Kinesiol grades the evidence strong — GRADE A — against an effect on muscle strength. The 2024 meta-analysis in Sports Medicine (Bischof and colleagues, 54(11):2865–2888) did find a statistically significant effect on maximal strength, and the number is SMD 0.19, on evidence its own authors rated low certainty. That is the ceiling of this literature.
Both halves of that are worth keeping, because they are usually quoted one at a time by people who want opposite things.
A significant result is not the same as a result worth paying for. An effect size of 0.19 is the sort of difference that needs a meta-analysis of nineteen trials and 768 people to become visible at all, and the review’s own authors graded the certainty of it as low. Quoting the number is more honest than either “collagen improves strength” or a flat “no effect” that the paper does not support.
The review that puts the strongest wording on it is the newer one. The same 2026 review this page cites for the one positive finding is also the paper that grades the evidence strong against a strength effect — both conclusions, from the same eight trials, in the same document. Splitting them is the misuse.
Is collagen a muscle-building protein?
No. Aussieker and van Loon, Med Sci Sports Exerc 2023;55(10):1792–1802, infused a tracer and took muscle biopsies from 45 people: 30 g of collagen after a resistance session did not raise muscle connective protein synthesis above placebo, while whey in the same trial raised myofibrillar synthesis. Collagen is not a muscle-building protein and TARE does not sell it as one.
That design is the strongest one in the field — a labelled amino acid infused into the bloodstream and the muscle itself sampled with a needle, rather than a number inferred from a scan. And the same group kept testing it.
- The 2025 replication. Forty-five men and women, 30 g of collagen against 30 g of the same amino acids supplied in free form against a placebo. Neither the collagen nor the free amino acids raised myofibrillar or connective protein synthesis in the rested or the exercised leg. Med Sci Sports Exerc 2025;57(11):2394–2408.
- Against pea protein, in older men. McKendry and colleagues, Am J Clin Nutr 2024;120(1):34–46, added 50 g a day of whey, pea or collagen to a controlled diet for 31 men over 72. Whey raised integrated muscle protein synthesis. Pea raised it. Collagen did not move it.
The second one is the sentence that matters. Collagen did not merely lose to whey, which is a premium protein people expect to win. It lost to pea, which nobody markets as premium, in a design where the only variable was which powder went into the same diet.
Does collagen reduce injuries?
No trial has reported it. Searches for collagen or gelatin against injury incidence return no randomised controlled trial with injuries as a measured outcome. That is an absence rather than a null result, and it is a weaker position than a null. Papers in this field do speculate about injury risk in their closing sentences without having measured it, which is how the claim gets into circulation.
This is the claim the category leans on hardest and it is the one with nothing underneath it at all. A marker that rises in blood and a tendon property that changes with training are both real findings. Neither of them is a count of injuries in a training population, and no amount of mechanism converts one into the other.
The softened forms are the same claim and this site does not use them either — not for resilience, not for staying available, not for durability. They mean the same thing while being harder for a reader to check, which is the only reason they exist.
What is the study the whole category rests on?
Shaw, Lee-Barthel, Ross, Wang and Baar, Am J Clin Nutr 2017;105(1):136–143. Eight healthy men, three days, 5 g against 15 g of vitamin-C-enriched gelatin one hour before six minutes of rope-skipping. The 15 g arm roughly doubled serum PINP. PINP is a marker measured in blood, not a tendon, and eight men is the entire human sample.
Nearly every dose recommendation in this category, including the one on this site, traces back through that trial. It is worth being precise about what it did and did not contain.
- Eight people. All healthy males. No women, no older adults, no injured population.
- Three days. No adaptation window, and no tendon, ligament, strength or injury outcome was measured in the human arm.
- The improved-mechanics result is in a dish. Engineered ligament constructs treated with the subjects’ serum, not human tissue.
- The follow-up did not replicate it cleanly. Lis and Baar, Int J Sport Nutr Exerc Metab 2019;29(5):526–531, ran the same idea in ten men: the marker rose about 20% with gelatin and with hydrolysed collagen, and the variability was too large for any treatment to reach significance. The gummy form did nothing.
That last line shares an author with the 2017 paper. When the people most invested in a finding publish the run where it did not reach significance, that is the sort of thing a reader should be told by the brand selling the powder rather than by somebody else.
What does the evidence actually support?
Tendon cross-sectional area and stiffness, and only those. The 2026 systematic review in J Funct Morphol Kinesiol — eight randomised trials, 257 people — grades that evidence strong at 15 to 30 g a day, in trials that every one of them paired the powder with resistance or plyometric training. Three of the four trials measuring area found a larger increase with collagen, and stiffness separated only in the trials dosing 15 g or more.
Two words carry that sentence: with loading. Every structural finding in the review came out of a trial where people were training hard, and the powder is the smaller half of that intervention. It is why the dose on this site is tied to a session rather than to a day, and why there is no rest-day instruction.
The cleanest single trial underneath it declares no conflicts of interest. Nulty, Phelan and Erskine, Eur J Sport Sci 2025;25(4):e12281 — 20 recreationally active men with a mean age of 47, 12 weeks of progressive lower-limb training, 30 g of collagen with 50 mg of vitamin C against maltodextrin carrying the same vitamin C. Patellar tendon cross-sectional area, stiffness and Young’s modulus all rose more in the collagen group. Maximal torque, rate of torque development, ten-rep max and muscle thickness rose with the training and showed no group difference at all. Nine men took the collagen.
So the honest version of the positive finding has four qualifiers attached, and none of them is optional.
- It is an imaging measurement, not an outcome. Cross-sectional area and stiffness are pictures of a tendon. Nobody in these trials measured an injury, a performance result or a pain score that TARE could stand behind.
- The evidence base is 95.7% male. Eleven women appear in the whole of the 2026 review’s 257 participants. The one trial in professional female athletes — Lee, Robshaw and Erskine, Exp Physiol 2025;110(11):1551–1560 — randomised eleven people, and six of them took the collagen.
- The trials are small and short. A mean of about 32 people each, three to fifteen weeks.
- The trial that ran 15 g for 15 weeks and imaged the tendon by MRI found nothing at all. That is the section above, and it does not stop being true because this section exists.
Our own dose sits on the null side of that range
This is the part of the section that costs us something, so it is printed rather than left for a reader to assemble. Of the individual trials named on this page, the two that separated from placebo on tendon stiffness both dosed 30 g — the 12-week trial in middle-aged men above and the 10-week trial in professional women footballers. The trial that dosed 15 g and imaged the tendon by MRI separated from placebo on nothing. Fifteen grams is the figure this site recommends.
The reason it is 15 g and not 30 g is that 15 g is the dose in the 2017 trial that established the timing, and that 30 g has never been shown to add anything where the measurement was a muscle biopsy rather than a scan. A 2024 crossover of eight middle-aged men — Nulty and colleagues, Am J Physiol Endocrinol Metab 2024;327(5):E668–E677 — did find the blood marker rising further at 30 g than at 15 g, and a blood marker is not a result. We did not go to 30 g. That is a judgement rather than a finding, and this is what it looks like when it is shown to you instead of rounded off.
Who paid for the favourable half of this literature
This is the part that changes how the rest of the page reads, and it is not a conspiracy claim. It is a funding line, printed where the result is printed.
The most widely cited framing document in the field — Holwerda and van Loon, Nutrition Reviews 2022;80(6):1497–1514 — is the paper that proposed collagen as a glycine and proline source for connective tissue. Its listed funders are PB Leiner and the Tessenderlo Group, a gelatin and collagen-peptide manufacturer. It is a narrative review, so it carries no systematic search and no pooled estimate, and it framed the idea as a hypothesis to be tested. The same authors then tested it in the tracer trials in the muscle section above, and it was not confirmed.
The pattern repeats one tissue over. The 14-week trial that grew the Achilles tendon on 5 g a day — Jerger and colleagues, Scand J Med Sci Sports 2022;32(7):1131–1141 — lists Gelita AG, a collagen manufacturer, as its funder. In that same trial, tendon stiffness and muscle strength rose equally in both groups. The manufacturer’s own study found no difference on the mechanical property that matters.
None of that makes a funded result false. It does mean the funding line belongs in the same sentence as the result, and the reason this page prints it is Myung and Park: in the one literature where somebody actually ran the comparison, splitting by funder was enough to make the effect disappear.
Who is collagen for, then?
Someone who already trains hard and wants to load a tendon, at 15 g with 50 to 100 mg of vitamin C, 30 to 60 minutes before the session, on training days. Not someone buying it for skin, hair, nails, strength, muscle or injuries, because on those the evidence is either against it or absent. And not someone hoping to replace a protein powder.
The fair criticisms are worth putting in our own words rather than waiting for them. Collagen carries zero tryptophan. It is low in leucine. Its DIAAS is effectively zero, which is the formal way of recording that it is a poor protein for the purpose most protein powders are bought for. And “it is hydrolysed gelatin” is largely true — the 2017 trial that started all of this used gelatin, not peptides.
None of that is answered by a better sentence. It is answered by not selling collagen as a protein supplement, which is why the comparison against whey is its own article and why the answer in it is to buy whey.
What is left after all of that is a narrow, real thing: a pre-training dose, with a cofactor, tied to a session, supported by a systematic review that also says it will not make you stronger, and contradicted by the best-imaged trial at the same dose. If that sounds like a smaller product than the category sells, it is. It is also the version that survives being checked.
Where our own tub sits
Under Load is collagen Type I and III, bovine, one ingredient, 350 g, with a 5 g scoop. The Supplement Facts panel declares 10 g, two scoops, 35 servings. The protocol on this site asks for three scoops before training, which is a directions statement and not a serving declaration — and at that dose the tub holds 23 whole doses, not 35. The arithmetic is done in public on how much collagen for tendons.
The certificate behind it is a real test of a real 350 g lot and it is not an independent one. ViaCore Nutrition, El Reno, Oklahoma, is the manufacturer’s own laboratory — lot 121125450, analysis dated 12/22/2025, protein 9.8 g per 10 g serving against a 9 g specification. That is a weaker check than an outside laboratory and a reader should hear it from us. What the document does and does not cover is transcribed row by row on Testing & Certification.
We audited our supplier and two products did not survive it. They were retired rather than quietly restocked, and the reasoning was published rather than buried. That is the same instinct as this page: the sections above are what the evidence says, including the parts that cost us the sale.
More in this cluster: collagen against whey, and why we tell you to buy whey, how much collagen for tendons, what a branded collagen peptide is, and what it is not, and what a certificate of analysis proves, and what ours do not. Or read the documents themselves on Testing & Certification.