The research behind LookFlex

How we choose and verify sources

We would rather say less than say something we cannot back up. So the rule for this site and the app is simple: a factual claim about faces, attractiveness, skin, sleep, nutrition, training, habits or our own methods must point to a source in this list, or it must be written plainly as guidance, styling convention or our own model.

• What we look for. Pooled analyses (meta-analyses), randomised trials, official guidelines and large studies come first. Where only small or early studies exist, we use them, label them as limited and say what they cannot show.
• How we check them. Every DOI is checked against the Crossref registry with a script, scripts/verify-research.mjs, and sources without a DOI are checked to make sure the page loads. A source is not added until it passes. That confirms the paper exists and that the title, year and first author match its record. It does not replace reading the paper, so follow the link if a finding matters to you.
• How we write them up. Each finding is a short plain-language paraphrase in our own words, not a copy of an abstract. We try to keep every caveat that matters, such as small samples, observational designs and industry funding.
• What we do when evidence is thin. We say so. For some popular claims, such as tongue posture reshaping adult bone, we found no reliable evidence and do not present them as fact.

Found a mistake or a better source? Write to [email protected] and we will look at it.

How to read the evidence labels

Each source carries one label, which describes how much weight that source can carry on its own:

• Strong evidenceLarge, well-designed or pooled studies, or an official guideline. A claim can lean on this source.
• Moderate evidenceReasonable studies with some limits, such as small samples, a single lab or an industry sponsor. Useful, not conclusive.
• Limited evidenceSmall, early or indirect studies. Interesting, not enough to rely on.

The label is our judgement of the source, not the journal's. It can also mean different things for different questions: a strong study can show that an effect is small, as with face width and aggression.

What this research cannot tell you

• It does not validate LookFlex's scores. No study here tested LookFlex. The reference ranges, measurement margins and scores come from our own calibration, described on the methodology page, and we have not had them independently audited. The measurement-error sources explain the method we use for margins, not how precise LookFlex is.
• It is mostly about perception. Most appearance research asks how strangers rate photos. That tells you how people respond, not what is healthy or good for you, and tastes differ between people.
• Samples are limited. Many studies use small groups, students or mostly European faces. We flag this where it matters, and we test our own tools for fairness across skin tones rather than assuming it.
• It is not medical advice. See the medical disclaimer, and speak to a doctor or dermatologist about health concerns.