Who to trust: evaluating fitness information sources

Nobody can evaluate every claim from first principles. You have to rely on sources, which makes choosing sources the actual skill.

Signals of a good source

They express uncertainty. The single best indicator. Someone who says "the evidence here is weak", "this is my opinion rather than a finding", or "we do not know" is engaging honestly with a field where most questions are unsettled.

Total confidence about a contested question is a warning sign, not a credential.

They change their position. Anyone who has been producing content for a decade and holds identical views to their first year either was extraordinarily prescient or is not updating. Public corrections are a strong positive signal.

They cite specific research and describe its limitations. Not "studies show" — an actual study, with the population, duration and outcome measure noted. Someone who tells you a finding came from twelve untrained undergraduates over six weeks is being straight with you.

They distinguish between magnitude and existence. A great deal of bad advice consists of true statements about effects too small to matter. Good sources tell you how big something is, not just that it is real.

They disclose conflicts. Selling a programme, a supplement or coaching is not disqualifying. Failing to mention it while recommending the category is.

They agree with other credible sources on the fundamentals and disagree at the margins, which is what a healthy field looks like.

Signals of a poor source

Secrets and revelations. "What the industry does not want you to know." There are no secrets in exercise science; the papers are public.

Single-study certainty. A new paper contradicting a body of work is a reason for interest, not for reversing your practice.

Physique as the argument. Being very muscular demonstrates genetics, consistency and possibly pharmacology. It does not demonstrate that the person understands why they got that way, and elite athletes are frequently poor sources of general advice because their experience is unrepresentative.

Everything is a hack, a mistake, or a myth. Content structured entirely around contrarianism is optimising for attention.

Attacking others rather than evidence. Fitness has an unusually adversarial culture and the people producing the most valuable work are generally not the ones spending their time on it.

Selling the thing they are recommending, without saying so.

Anything absolute. "Never do X." Very few things in training are never appropriate.

Credentials, honestly assessed

Relevant qualifications — a research degree in exercise physiology, a physiotherapy or dietetics registration, an established coaching record with a variety of athletes — shift the prior somewhat.

They are not sufficient. There are qualified people producing poor content and unqualified people producing excellent content, and the behavioural signals above are more diagnostic than the letters.

Certifications in particular vary enormously in rigour, from serious multi-year programmes to weekend courses.

The three-source habit

Before acting on a training or nutrition claim that would meaningfully change what you do, check whether independent credible sources agree.

Where they converge, confidence is reasonable. Where they diverge, the honest answer is that the question is unsettled — and unsettled questions are usually unsettled because the effect is small, which means the decision matters less than it feels like it does.

The proportionality filter

The most useful single question when evaluating any claim: how much would this change my results if it were completely true?

An enormous amount of content concerns variables that, at best, account for a few percent. Rest interval optimisation, exercise order, supplement timing, specific rep ranges.

Meanwhile the variables that dominate — showing up consistently for years, progressive overload, eating enough, sleeping enough — are boring and generate almost no content because there is nothing new to say about them.

If a source spends most of its output on the small variables, that is a signal about what it is optimising for.

Your own data

Ultimately the most relevant evidence about what works for you is your own tracked record over months.

Research gives you the starting point and the prior probability. Your training log tells you what actually happened. When they conflict for you specifically, your log wins — provided it is a log and not a memory.