How we read the evidence

Human studies only — read first-hand

Every claim traces to a peer-reviewed study in people. No animal studies. No cell-culture results standing in for a human outcome. And no abstract-only citations: the full text has to be accessible and read first-hand, because the abstract often states an effect the methods and results do not support. Put plainly: we build only on evidence you can open and read yourself.

The primary evidence window is 2020–present, with older landmark trials used when newer human evidence is thin. Sources are open-access journals and repositories whose full text anyone can verify — so a citation is a link you can open, not a paywall you have to trust.

Food and methods — one bar, two gates

This site covers two kinds of thing: what you eat, and what you do — proven lifestyle methods like fasting, heat and cold. Both are held to the same human-evidence bar, through the gate that fits each.

The food gate. A food claim — "eat this amount of this food to get this result" — is published only when a human trial measured that result using a food-achievable dose: a normal serving of the whole food, or a freeze-dried or powdered whole food at a food-equivalent amount. When the only evidence is a concentrated extract or a supplement, the page says what the compound does mechanistically and stops short of promising food will reproduce it.

The method gate. A method claim — "do this to get this result" — is published only when a human trial actually ran that method and measured the outcome, and it is rated separately for each outcome. Cold water might be reasonable for short-term mood yet unproven for fat loss; the page says so plainly rather than giving the method a single blanket verdict.

In both gates the effect is anchored on a human trial, not a meta-analysis — reviews pool mixed forms, doses and protocols, so they corroborate a finding but never set the number or the verdict.

No supplements recommended — and why that is the point

We recommend no supplement, extract or supplement brand, and will never sell one — that promise, never to sell, is permanent. That is not a disclaimer bolted on; it is the standard. In the UK, US and most markets supplements are regulated as food, not assessed for effectiveness before sale, and a paid supplement recommendation is exactly the conflict of interest that makes most longevity advice untrustworthy. We cover food and lifestyle methods only. Where a supplement is the only evidence that exists, we report it honestly and leave the decision where it belongs — with you and a clinician.

Our default is no — but it's a considered position, not a slogan. Most supplements aren't assessed for effectiveness before sale, and their quality can be inconsistent, so we cover food and proven methods instead. We hold this the way a scientist should: provisionally. If a specific supplement ever met the same standard we apply to everything else, we'd say so and change our view. And the ground may shift — in this new AI era, better research, formulation and manufacturing could make parts of the supplement industry more consistent and better evidenced. If that happens, we'll follow the evidence.

Every result names its population

A finding belongs to the people it was measured in. So each claim states its population in plain language — "in overweight adults," "in postmenopausal women," "in adults with high blood pressure." A result from a clinical group is never dressed up as if it applies to everyone, and when a benefit has only been shown in one population, the page says the healthy-person evidence is missing rather than implying it.

Where the evidence stops

Some of the most-shared longevity claims have no qualifying human study behind them. When that is the case, the honest answer is "not supported" — and that answer gets its own page, stated as plainly as any protocol. A page that tells you what a food or method cannot do is doing the same job as one that tells you what it can: keeping the standard the same in both directions.

Evidence vs. experience

We rate the evidence for a specific claim, in a specific outcome — not a whole topic, and not your personal result. Three things follow from that, and they matter:

• A weak or "not supported" rating means the human trials have not shown the effect. It does not mean the thing cannot work for you. Absence of evidence is not evidence of absence.
• People vary. A result measured across a group is an average; your own response can differ, for reasons a trial was never designed to capture.
• How something makes you feel is a real outcome, even when no trial has measured it. We say where the trial evidence is thin and, separately, where people report a genuine effect — without dressing one up as the other.

This is the deliberate limit of the method: we rate trial evidence, because it is the most reliable guide there is — but it is not the last word on what might help you. Where something is worth trying, treat it as a personal experiment rather than a proven protocol, and mind any real risks.

How these answers are researched

We think you should know how these answers are made, because the honesty of the method is the whole point. The reading and extraction is done with substantial help from AI: it searches the approved full-text sources, reads the papers, and pulls out the specific numbers, populations and caveats. What AI does not do is decide what is true. The standard on this page is fixed by us, every answer is checked against it, and a person reviews and approves each page before it is published. The tool helps us read at scale; the judgement, and the responsibility, are ours.

The study a verdict rests on — the anchor — is retrieved and read in full, first-hand, and its figures are read back against the source before they reach the page. Corroborating reviews and meta-analyses may be read at summary level; as the standard above says, they can support a finding but never set the number or the verdict. So when a page cites a figure, it comes from a full paper someone opened — not an abstract.

On top of that standard, each answer runs a set of checks built to catch the two ways this kind of research goes wrong: inventing something that is not in the paper, or stretching a real result further than the study allows. Every claim has to name a specific human study with an openable full-text link. The numbers on the page are matched back to the numbers in the paper. Verdicts are held to fixed rules so nothing is quietly rounded up — it is why beetroot is rated strong for systolic pressure only, blueberries moderate, and some popular claims "not supported." Hard rules block particular overreaches: joining two separate findings into a claim no trial tested, turning an "associated with" into a "causes," or presenting a supplement result as though it came from food. And the population, and the limits, are stated on every page.

No process, human or automated, is perfect, and we will not pretend these checks make the pages infallible — that would be its own kind of overclaim. What they do is make errors rarer and easier to catch. What catches the rest is the same thing that catches any publisher's mistakes: a dated review, an open correction, and readers who tell us when something is off. That is why the corrections policy below matters as much as the method above.

Corrections & updates

Getting something wrong is not the failure; hiding it is. When an error is found — a mis-stated figure, a study that does not support the claim, a verdict overtaken by newer evidence — it is fixed openly. Material corrections to a fact, figure or verdict are marked on the page with the date and a note of what changed, never quietly overwritten. Minor edits, like a typo or clearer wording, are made without a note.

Every page also carries a "last reviewed" date, and when new human evidence moves a rating, the "what would change this verdict" line becomes the record of why it moved. If you think something here is wrong, please tell us — with the page and a source, if you have one — and we will check it against the evidence and correct it if it stands.

Not medical advice

Everything here is educational information drawn from published human research. It is not a diagnosis, a treatment plan, or a substitute for your own clinician, who knows your history and medication. Use it to ask better questions, not to replace medical care.