Evidence intelligence
Every claim
has an anatomy.
Paste a health or biology claim. MyAtomy traces the evidence, evaluates the research, and shows you what the science actually supports.
No account required · Free to use
Try one
A verdict isn’t enough. See how the evidence is built.
Every report takes the claim apart into the things that actually decide whether it holds: how much human evidence exists, whether the studies agree, how good they are, what contradicts them, and where the whole thing breaks down.
Evidence anatomy · illustrative layout
Six dimensions01 · Evidence strength
How much weight the human evidence actually carries.
Not a probability and not a p-value — a heuristic read on the size, quality and consistency of what has been demonstrated in people.
02 · Biological mechanism
Could it work, and does that matter?
A plausible pathway is not a result. Mechanism and demonstrated outcome stay in separate columns, because conflating them is how most health claims go wrong.
03
Human evidence
Moderate
04
Consistency
Limited
05
Study quality
Limited
06
Contradictions
Conflicting
07 · Limitations
Where the claim breaks
- Studied in athletes, claimed for everyone
- Measured a marker, not an outcome
- One trial, never replicated
- Doses no one would realistically take
08 · Source trail
26Every source it read, listed and linked.
Records come from Europe PMC and Crossref, not from the model’s memory. If nothing solid was found, the report says so instead of filling the gap.
- [1]Meta-analysis14 trials pooled
- [2]RCTRandomised crossover
- [3]CohortProspective, n = 4,100
- [4]AnimalRodent model
Three stages. No shortcuts.
The pipeline is deliberately boring: structure the claim, go and read, then report what is there — including when the answer is that nobody really knows yet.
- 01
Submit the claim
Paste a sentence, a link, a DOI, a paper or a screenshot. MyAtomy restates it as a precise, testable question and breaks it into population, exposure, outcome and mechanism.
- 02
MyAtomy traces the evidence
Several search angles run against biomedical literature databases. Results are deduplicated and ranked on relevance and design before anything is synthesised.
- 03
See what survives scrutiny
The evidence is weighed against the claim as written — what it supports, what it does not, where the studies disagree, and how confident anyone should be.
Not all evidence is the same weight.
Different claims need different evidence. A question about a biological pathway can be settled in a dish. A question about whether people live longer cannot.
- 01
Systematic reviews & meta-analyses
Strongest priorPool many studies to estimate an overall effect. Powerful when the underlying trials are good, misleading when they are not.
- 02
Randomised controlled trials
Strongest priorRandomisation is the cleanest way to separate an effect from the people who happened to receive it.
- 03
Controlled human trials
Strong priorNon-randomised but controlled human work. Useful, more open to confounding and selection effects.
- 04
Observational evidence
Strong priorCohorts and case-control studies show association at scale. They rarely establish causation on their own.
- 05
Case studies & series
Moderate priorDetailed accounts of a few individuals. Good for generating hypotheses, weak for general conclusions.
- 06
Animal research
Moderate priorEstablishes biological possibility. Doses, physiology and lifespans differ enough that translation often fails.
- 07
In-vitro & mechanistic work
Weakest priorCells in a dish show a pathway can operate. It does not show the pathway matters in a living human.
The same sentence, taken apart.
A claim arrives as one confident line. Underneath it are five separate questions, and they rarely all have the same answer.
The claim
“X has been scientifically proven to improve Y.”
Confident, quotable, and impossible to check as written. It names no population, no dose, no comparison and no timeframe.
Each phrase hides a question
- “scientifically proven” — How strong is the evidence, really?
- “improve” — By how much, measured how?
- “Y” — In whom, and does it cause it?
The anatomy
Example layout- Evidence
- Mixed
- Confidence
- Moderate
- Human evidence
- Limited
- Mechanism
- Plausible
- Causality
- Not established
Credible studies point both ways.
Enough literature to judge, not enough to settle.
Small, short trials in narrow populations.
A pathway exists. That is not a result.
Association, not demonstrated cause.
An illustration of the report layout, not a real analysis. Every live report is generated from literature retrieved at the moment you run it.
Science without
the spin.
Understand what a study found, what it didn’t, and how confident you should be.
Run an analysisBuilt for people who question the claim.
MyAtomy was created around a simple principle: before accepting or rejecting a scientific claim, inspect the evidence underneath it.
The premise
Science shouldn’t require a translator.
MyAtomy separates what a claim says from what the research can actually support. It is built to disagree with whoever is wrong — the person who submitted the claim, the source it came from, popular opinion, and the person who built it. Analyses are not tuned toward any individual’s views.
Founder
Research without another subscription.
Pay-as-you-go, per analysis. No monthly plan, no account, and nothing charged until you ask for a report.
Every analysis is free right now. The prices below are proposed for when checkout opens.
Quick Check
One focused claim
A single claim traced against the strongest available literature, with a verdict, mechanism read and source trail.
- Claim normalisation
- Evidence verdict + confidence
- Mechanism assessment
- Core source trail
Free today$0.50Proposed · per analysis
Paper Anatomy
Research-paper breakdown
A single paper dissected: design, population, effect size, whether the conclusion matches the data, and what headlines tend to get wrong.
- Design + methods read
- Effect size and significance
- Strengths, weaknesses, confounders
- Conclusion-vs-data check
- What headlines get wrong
Free today$1.00Proposed · per analysis
Deep Evidence Scan
Most thoroughExpanded literature analysis
Wider literature sweep across more query angles, with contradiction hunting and a fuller methodological read.
- Everything in Quick Check
- Broader literature sweep
- Contradiction analysis
- Study-quality assessment
- What would change the verdict
Free today$2.50Proposed · per analysis
Crypto payments coming soon
USDC · SOL on SolanaCheckout is not live. No wallet connection is required today, and nothing on this page can charge you.
No account, no card, no wallet.
FAQ
Questions worth asking.
Claims about health, biology, medicine, nutrition, genetics and physiology — the sort of statement that research could in principle support or undercut. It handles a single sentence, a paragraph of marketing copy, a paper, or a screenshot of a post. It is not built for questions about politics, law, finance or history, and it will say so rather than improvise.
It searches Europe PMC, which indexes PubMed, MEDLINE, preprint servers and full-text open-access articles, and uses Crossref to resolve DOIs. Several search angles run for each claim, results are deduplicated and ranked, and the model may only cite records that were actually retrieved. It cannot cite from memory.
No. It explains what published research shows. It does not diagnose, does not recommend starting, stopping or changing any medication or dose, and does not account for your own history. If you are deciding something about your own care, the report is background reading for a conversation with a clinician, not a substitute for one.
Yes, and it switches to a different report when you give it one. Paper Anatomy covers the design, population, methods, effect size, statistical versus practical significance, strengths, weaknesses, confounders, whether the authors' conclusion matches their own data, and what coverage of a paper like this typically overstates. Paste a DOI, a PubMed link or the abstract, or upload the PDF.
Yes. Upload a screenshot of a post, an infographic, a chart or a supplement advert and MyAtomy transcribes what it can read and lists the testable claims it found. You pick which one to dissect. Text recovered from an image can be partial or misread, so the transcription is shown to you rather than treated as fact.
Confidence describes how much the retrieved literature can settle the question — quality, consistency, how directly the studies address the claim, and how much of the relevant work appears to have been found. It is separate from the size of any effect. A large effect shown once in a small trial gets low confidence; a small effect replicated across good trials can get high confidence.
A 0-100 heuristic summarising how much weight the human evidence carries for the claim as written. It is a reading aid, not a statistic: it is not a probability, not a p-value, and not a clinical measure. Two people could reasonably disagree about a given number. The written report is the substance; the score is a shorthand.
Yes. Every analysis is free today and nothing on the site can charge you. There is a limit of 5 analyses per browser per day so the service stays available for everyone. Pay-as-you-go pricing is planned for later, with no subscription.
No. There is no sign-up, no sign-in and no password. Your recent reports are stored in your own browser, on your own device, and clearing your browser data clears them.
When checkout opens, you will pick an analysis, send the exact amount in USDC or SOL on Solana, and the analysis will run. No subscription, no stored card, no account. It is not live yet, and no wallet connection is requested anywhere in the product today.