Is a single PubMed hit enough proof?

No. PubMed has over 35 million entries — including case reports, animal studies, hypothesis papers, and weak observational studies. A single hit proves nothing. Look for meta-analyses or systematic reviews on the same topic.

What are red flags in a biohacking claim?

Small n (< 30), animal model as main evidence, no conflict-of-interest statement, effect only reproduced in the manufacturer's own studies, p-hacking suspicion (multiple endpoints without Bonferroni correction), no placebo comparison, short trial duration for chronic endpoints.

How do I tell effect size from statistical significance?

Statistically significant (p < 0.05) only means: the effect is probably not random. Says nothing about size. Example: 'significantly lowers LDL' might mean −2 mg/dl — statistically significant, clinically irrelevant. Look for absolute numbers, effect sizes (Cohen's d), number needed to treat (NNT).

Why are animal studies often misleading?

Mice live 2 years, humans 80. Doses are usually scaled per kg body weight, which doesn't translate pharmacokinetics 1:1. Cancer studies in mice replicate in human trials only ~5-10 % of the time. Animal studies are hypothesis-generating, not proof.

Where do I find serious supplement reviews?

NIH ODS Fact Sheets (ods.od.nih.gov) — neutral, regularly updated, with evidence rating. Cochrane Reviews (cochranelibrary.com) — gold standard for meta-analyses. Examine.com — commercial but transparently referenced. For supplement reviews with industry ties: be doubly cautious.

How do I use BiohackingAI for fact-checking?

Ask a concrete question (e.g. 'Does curcumin reduce inflammation markers in healthy adults?'). Our AI searches PubMed live, weights by evidence level, and returns cited studies as clickable sources — no hallucinations, no invented study IDs. When in doubt it says 'data are limited' instead of making something up.

What if an influencer cites a study?

Read the study itself, not the influencer summary. Effect sizes get inflated 10x, study types get upgraded (animal → 'study'), or conflicts of interest are concealed. If the influencer sells the product the study tests: probability of honest portrayal drops sharply.

Biohacking fact-check: 7-step guide

Q: What are red flags in a biohacking claim?

Small n (< 30), animal model as main evidence, no conflict-of-interest statement, effect only reproduced in the manufacturer's own studies, p-hacking suspicion (multiple endpoints without Bonferroni correction), no placebo comparison, short trial duration for chronic endpoints.

Q: How do I tell effect size from statistical significance?

Statistically significant (p < 0.05) only means: the effect is probably not random. Says nothing about size. Example: 'significantly lowers LDL' might mean −2 mg/dl — statistically significant, clinically irrelevant. Look for absolute numbers, effect sizes (Cohen's d), number needed to treat (NNT).

Q: Why are animal studies often misleading?

Mice live 2 years, humans 80. Doses are usually scaled per kg body weight, which doesn't translate pharmacokinetics 1:1. Cancer studies in mice replicate in human trials only ~5-10 % of the time. Animal studies are hypothesis-generating, not proof.

Q: Where do I find serious supplement reviews?

NIH ODS Fact Sheets (ods.od.nih.gov) — neutral, regularly updated, with evidence rating. Cochrane Reviews (cochranelibrary.com) — gold standard for meta-analyses. Examine.com — commercial but transparently referenced. For supplement reviews with industry ties: be doubly cautious.

Q: How do I use BiohackingAI for fact-checking?

Ask a concrete question (e.g. 'Does curcumin reduce inflammation markers in healthy adults?'). Our AI searches PubMed live, weights by evidence level, and returns cited studies as clickable sources — no hallucinations, no invented study IDs. When in doubt it says 'data are limited' instead of making something up.

Q: What if an influencer cites a study?

Read the study itself, not the influencer summary. Effect sizes get inflated 10x, study types get upgraded (animal → 'study'), or conflicts of interest are concealed. If the influencer sells the product the study tests: probability of honest portrayal drops sharply.

Direct answer

In seven steps you can classify any biohacking claim as proven, plausible or BS in 5 minutes: 1) check study type (meta-analysis > RCT > observation > animal > anecdote), 2) check sample (n < 30 is a pilot, not proof), 3) read effect size, not p-value, 4) scan conflict of interest, 5) look for reproducibility (multiple independent groups), 6) go to PubMed directly, don't trust the influencer summary, 7) for health decisions, Cochrane or NIH ODS as gold standard.

The 7 steps in detail

Step 1 — Identify evidence level

The scientific world has a hierarchy:

Meta-analysis / systematic review — synthesizes many RCTs, highest level
Randomized controlled trial (RCT) — gold standard of single studies
Observational / cohort study — can show association, can't prove causation
Case series / case report — few patients, descriptive
Animal study / mechanism — biologically plausible but not human
Expert opinion / anecdote — weakest level

A claim based on a single mouse study has a completely different weight than one based on a Cochrane meta-analysis of 30 RCTs.

Step 2 — Check sample size

n < 30 = pilot, n < 100 = hint, n > 1000 = solid signal (for moderate effects). For small effects (e.g. cognitive outcomes) you often need n > 500 for statistical power. A study of 12 participants on lion's mane says almost nothing — even with p < 0.05.

Step 3 — Read effect size, not p-value

Statistically significant means: probably not random. Says nothing about size. Practically useless:

"Significant LDL reduction" — if that's −2 mg/dl from a baseline of 140: clinically irrelevant.
"Significantly improves cognitive performance" — 0.3 points on a 100-point scale: measurable, but not noticeable.

Look for absolute numbers, effect sizes (Cohen's d: 0.2 small, 0.5 medium, 0.8 large), number needed to treat (NNT) for clinical endpoints.

Step 4 — Scan conflict of interest

Who paid for the study? Who analyzed the data? Who profits from the result? Industry-funded studies are statistically 4-6× more likely to report positive results. That doesn't make them automatically wrong, but they need independent replication.

Most serious journals require a "Conflict of Interest Statement" at the end of the study. Missing or evasive: skeptical.

Step 5 — Look for reproducibility

A claim that exists in only a single study from a single lab is a signal — not proof. A lever is proven only when multiple independent research groups have replicated the result. For supplements and lifestyle interventions, many initially promising effects later fail replication (see "reproducibility crisis" in psychology and biomedicine).

Step 6 — Go to the source, not the influencer

Influencers have economic incentive to inflate effect sizes, drop caveats, and sell poorly supported claims as "studies show". Read the original. If it's not linked: strong red flag. If the influencer sells the product the study tests: be even more critical.

PubMed (pubmed.ncbi.nlm.nih.gov) is free and searchable. Abstracts are free. Full text for many papers is available via sci-hub.se (legal grey area, but functional).

Step 7 — Gold standard for health decisions

For decisions with health impact (supplements, drugs, diets):

NIH Office of Dietary Supplements Fact Sheets (ods.od.nih.gov/factsheets) — neutral, evidence-based, regularly updated
Cochrane Library (cochranelibrary.com) — gold standard for meta-analyses
UpToDate (clinical reference for physicians — paywall, but often accessible via libraries)

What doesn't count: Substack posts, YouTube summaries, Reddit threads, ChatGPT answers without sources.

Three common fallacies

"It worked for me" — n = 1, no control, no blinding, contaminated with placebo and expectation effects. Personal experience is valuable for hypothesis building, not for general recommendations.

"Thousands of years of traditional use" — argument from tradition. Bloodletting, skull trepanation, and mercury as medicine all had millennia-long tradition. Tradition proves nothing about efficacy.

"Studies in mice show…" — can be a hint but is not proof. 90+ % of promising animal studies don't replicate in human trials. Cancer mice have given us 50 years of cure hype without successful translation.

Methodology — how BiohackingAI automates this

The platform combines the seven steps into AI-powered search: each query triggers live PubMed retrieval, hits are weighted by evidence level (A+ to F), cited with clickable link to the primary source, and explicitly marked "data limited to…" where evidence is thin. No invented study IDs, no influencer claims taken uncritically.

How to fact-check biohacking tips? A 7-step guide