Where studies come from
Records come from PubMed, a service of the National Library of Medicine. We use Europe PMC to check each paper’s reuse license and Crossref to find the publisher’s page. Records may not reflect the latest NLM data.
How studies are chosen
Software searches PubMed for “photobiomodulation” and for related phrases, such as low-level light therapy, low-level laser therapy, red light therapy and near-infrared light therapy. A search runs every night, and every paper it finds is added, in any language and with or without an abstract. A few extra searches widen the smaller research areas: light therapy for mood and sleep, red light for myopia, light and hearing, and the safety of light and laser treatments. Each is a specific search, so papers that only mention light in passing stay out.
We leave out papers that PubMed marks as retracted, and items that are not research articles, such as comments, letters, editorials, news and corrections. We do not choose papers by their result, their authors or who paid for them.
How studies are sorted
Software places each paper in a research area by looking for keywords in its title, subject tags and abstract, and labels it a trial, review or study from the publication type PubMed gives. A paper has one main area and can also be listed in others when its title, subject tags or abstract clearly fit them, and systematic reviews and meta-analyses are also listed under Safety & Evidence. This is approximate, so a paper can end up in the wrong area. Papers that fit no area go to “Other”. Lists are sorted by publication date, newest first.
What you read on a paper’s page
The summary line is a short excerpt from the paper’s abstract, chosen by software, and is labeled as such. For recent papers we also show a short summary and a few facts, such as the number of participants and the wavelengths used. An AI model (Claude, from Anthropic) writes them from the abstract only, and software checks the numbers against the abstract. They are labeled as AI-generated and can still be wrong. We show the full abstract only when the paper’s license allows it. Light parameters and glossary terms are found in the text by pattern and may be incomplete. Some content may be generated or assisted by AI. See the disclaimer.
How to read each kind of research
Every paper page links here. The design of a study shows how far its result can be trusted.
How to read a trial
A trial tests a treatment on participants. The strongest kind assigns people to the real treatment or a sham by chance, and keeps both groups unaware of which they received.
- Was there a sham device, and could people tell it apart?
- Were participants and assessors blinded?
- How many people took part, and how many dropped out?
- Was the main result chosen before the trial began?
How to read a review
A review summarizes earlier studies. How much it can tell you depends on which studies it includes and how carefully it judged their quality.
- Did it search for studies in a stated, systematic way?
- Did it rate the quality or risk of bias of the studies it included?
- How different were the included studies (doses, wavelengths, people)?
- Does it say how certain the evidence is?
How to read a single study
One study rarely settles a question. Its design shows how far the result can be trusted. Case reports and small open-label studies can suggest ideas to test, but they rarely show that a treatment works.
- Was there a control or sham group?
- Were people unaware of which treatment they got?
- How many people took part?
- Do the authors call the findings preliminary or exploratory?
How to read a safety study
Safety studies look for harm rather than benefit. A lack of reported problems in a small or short study does not prove a treatment is safe.
- How many people were followed, and for how long?
- How were side effects recorded?
- Which doses and devices were tested?
What we do not do
- We do not grade, review or fact-check the studies, so a paper’s presence is not a sign of quality.
- We do not rewrite findings or give treatment advice.
- We do not rate or recommend products or devices.
The glossary and the guides to reading research below are there to help you judge the evidence yourself. Found a problem? Tell us.