What cooking does to collagen

Every whole-food collagen claim on this site runs through a pot. How long it simmers, how hot, and whether there's vinegar in it decide what actually ends up in the liquid — and the food-science literature on that is unusually good, because gelatin manufacturers have spent decades optimising exactly this.

Their findings redraw the folk picture in both directions — some things a long simmer is credited with it does not do, and one thing survives it better than anyone would guess.

Extraction plateaus, then reverses

The intuition is that longer means more. It doesn't, past a surprisingly early point.

A study optimising gelatin production from chicken feet reported no significant yield growth beyond about 2.5 hours of extraction. A response-surface study of chicken-feet collagen went further: in its fitted model, the coefficient for extraction time is negative, with the optimum at one to three hours. Longer runs recovered less. And a study tracking nutrient migration from chicken skeletons into soup across seven time points found total nutrient content peaked at 180 minutes and gained nothing at 210.

Prolonged high-temperature cooking also degrades what has already come out — protein oxidation rises and gel quality falls with extended heating.

So the collagen in a 48-hour stock came out in the first few hours. Hours three through forty-eight are doing something else.

Minerals are the exception

They keep climbing. In a controlled study of bone broths, calcium rose from 201 to 361 mg/kg and magnesium from 8 to 120 mg/kg going from a half-hour simmer to twelve hours, and cooking beyond eight hours extracted significantly more than shorter times.

Acid matters even more. Dropping the pH from 8.38 to 5.32 raised calcium extraction 17-fold and magnesium 15-fold. This is the real basis for the splash of vinegar, and it is a mineral effect, not a collagen one.

The caveat is that the starting point is low. Even at the top of that range, a serving delivers under 5 percent of daily recommended calcium and magnesium. The vinegar is doing something measurable to a quantity that doesn't matter much.

Simmering is a poor hydrolase

The most common overclaim is that long cooking breaks collagen down into the small peptides that supplements contain. Commercial collagen peptides are typically 2–5 kDa. Gelatin is tens to hundreds of kDa.

Heat does move the distribution, and further than the folk version of this argument allows. In bovine bone soup held at 121 °C, the fraction above 10 kDa fell by 21.4 percent across six hours. In chicken bone extracted at 130 °C, protein above 30 kDa was visible only in the 40-to-60-minute samples; by 90 minutes the extract was dominated by material under 10 kDa and between 10 and 30 kDa.

What heat does not do is reach the supplement range. Both of those are pressure retorts running 20 to 30 degrees above the 100 °C ceiling of a stockpot, and both leave the bulk of the protein an order of magnitude above 2–5 kDa. Chicken feet held at 100 °C for nine hours yielded plenty of collagen at an average of about 12 kDa, with roughly a fifth of it already under 5 kDa, but needed an added protease to bring the average to ~5 kDa; enzyme, not time, was what moved the molecular weight.

Home stock is gelatin. That is what the measurements describe.

Whether it matters is genuinely open, and the evidence leans towards less than you'd think. In a crossover study, intact gelatin raised blood Pro-Hyp and Hyp-Gly nearly as effectively as a hydrolysate did, and the authors concluded gelatin works as a functional food about as well. Free hydroxyproline absorption was lower from gelatin, so they are not identical — but the gap is smaller than the molecular-weight difference suggests.

What cooking destroys

Undenatured type II collagen — the UC-II class of supplement — works at 40 mg through immune tolerance, and that mechanism depends entirely on the triple helix staying intact so it can present native epitopes to gut lymphoid tissue. Denature it and the effect disappears; that was demonstrated in the foundational animal work, where heat-treated collagen did nothing while native collagen suppressed arthritis.

The isolated collagen triple helix denatures at about 41 °C, dropping to 35 °C once oxidised. Collagen inside intact tissue is more stable than that, and the reason has been measured: in a packed fibril the stretch of molecule that must be thermally activated at once is 26 residues, against 66 in a fibril swollen apart in acid. Neighbouring molecules hold each other in place, so a temperature measured on isolated collagen is a floor for what tissue does rather than an estimate of it.

Two consequences worth keeping. The first is that denaturation is a rate, not a threshold — at 47.4 °C the native collagen in one tissue fell tenfold in about twenty-one minutes, so temperature and time trade against each other and "what temperature does collagen denature at" has no single answer. The second is that a pot at 95 °C for two days is not a borderline case under any of these numbers.

One caveat on the numbers people quote here. The 60–75 °C figures that circulate in food writing for collagen shrinkage in intact meat could not be traced to a primary indexed measurement in mammalian tissue; searching the shrinkage- temperature literature returns leather chemistry and biomaterials scaffolds. The kinetic measurements above are what this entry stands on.

This is why "chicken feet contain type II collagen" is true yet does not carry the claim built on it. Whatever native type II is in the feet does not survive being cooked, so none of the UC-II evidence transfers to food. If you want that mechanism, the supplement is the route; cooking cannot get you there.

What nobody has measured

Worth stating plainly, because the absence is doing real work in every argument above. No published study has run SDS-PAGE, size-exclusion or GPC analysis on a domestic stock simmered for 24 or 48 hours. Every molecular-weight number in this entry comes from pressure-retort conditions or enzyme-assisted extraction, except one: the nine-hour 100 °C chicken-feet extract that the enzyme study measured before its enzyme had acted, which is the closest thing to a stockpot anyone has put through a column.

Nor has anyone measured what culinary-strength vinegar does to collagen extraction specifically. The acid studies use 0.3–5 percent acetic acid; a tablespoon or two per litre works out to roughly 0.05–0.1 percent — off by one to two orders of magnitude.

The direction of these effects is well established; the magnitude at real domestic cooking times is unknown.

What comes out of the pot, and what doesn't

The whole entry as a ledger. The presence column is about a domestic stock specifically — not a retort, not an enzyme-assisted extraction — which is why two of these rows are gaps rather than findings.

Note what the last two rows do not have in common. The small-peptide row is unmeasured — a domestic stock might be full of them, and the only reason nobody knows is that nobody has run the gel. The undenatured row is absent, and no future measurement will move it. Both look like bad news for a pot; only one of them is settled.