Methodology

The entries are written from evidence-acquisition reviews that Applied Wellness ran in Elicit. Each review searched the published literature on one question, screened what it found, and extracted what the included papers reported. These are the company’s own internal reviews: they are not independent of it, and they are not described here as systematic reviews to any outside standard.

So far there are two kinds:

• 31 topic reviews, each on one area of wellness, such as strength training, sleep duration or financial wellness. They hold 9687 source records.
• 7 option reviews, each on the named options within one kind of practice, such as breathing techniques or stretching methods. They hold 1153 source records.

Of all 10840 source records, 5166 were extracted from the full text, 753 from the abstract only, and 4921 were included when the results were screened a second time and have not been extracted. A paper two reviews both found is two records.

A source is shown by its study design and its year, for example “meta-analysis · 2021”, with its authors, its own title and its DOI, a link to the paper. Beside it, the app says how the source was read: for example “based on the abstract, not the full text”. No source carries a rating of any kind.

Some records are not research papers: protocols, preprints, conference abstracts, guidelines and the like. They stay visible, with a plain label and their DOI, and they are counted in no entry’s source count.

• Protocol / registration. A registered plan for a review or trial, not its results.
• Preprint. Posted by its authors before peer review.
• Conference abstract. A conference summary rather than a peer-reviewed paper.
• Peer-review report. A referee's report on a paper, not the paper itself.
• Guideline / position stand. A recommendation derived from evidence, which is a step further from the data than the evidence itself.
• Book chapter. A chapter in a scholarly book rather than a journal paper.
• Type not recorded. Our record does not say what kind of publication this is, so it is counted in no source count. That is a gap in our data, not a judgement about the research.

A claim is shown by the facts its sources stated, and nothing else:

• how many sources there are, and of what design;
• whether they agree at that exact claim;
• what each source measured, and how, in its own words;
• for a review, what it says it includes: its designs, and its studies and participants, as it states them;
• the certainty a source published about its own evidence, quoted and attributed to that source, or the plain statement “No source states a certainty rating.”

Where the sources disagree, the entry says so first: the badge adds “sources disagree”, the claim’s first sentence leads with the disagreement, and the evidence panel shows which sources found what. You can hide those claims with the filter switch “Hide claims where the sources disagree”; it is off unless you turn it on.

The sentences an entry carries are written to the strength of the source and never more, and a hedge in them is the source’s own. The app does not assess risk of bias. It records what a source did and what it concluded.

Earlier versions of this app showed a letter grade on each entry. We removed it: the reviews behind the entries did not do the risk-of-bias assessment a grade requires, and a letter hides the reasoning this app exists to teach.

The app keeps three things apart, and its copy never blurs them:

• Not yet recorded. We hold the source, but our record does not yet hold that fact about it. Where the direction a source reports at a claim is not yet in our record, the Agreement line says for how many sources it is recorded and that it is not yet recorded for the rest.
• The sources are silent. The sources do not state it. On a routine’s safety lines this is labelled “Source silent”, which is not the same as saying the routine carries no risk.
• Read from the abstract. The review read the paper’s abstract and did not extract from its full text. The source is labelled “based on the abstract, not the full text”.

The app writes structure: navigation, labels, groupings, the mapping of practices to the requirement they fill, calendar scaffolding, and the sessions inside a plan. Each is labelled as the app’s. A step whose sets, reps or time no source states is labelled “App-designed prescription”; a safety line the app wrote is labelled “Product safeguard (app-authored)”; a frequency or time of day no source states is labelled “planning default, not from the research”.

The app does not write findings, doses, comparisons or safety claims. Each practice carries the dose its sources studied. You can schedule any length you like; below the studied dose the label reads “This is the studied dose. Benefit at a shorter dose is not established here.” Where the research states no minute count, none is invented.

• It never rates evidence, and it never ranks practices.
• It never orders a list by any evidence fact. Lists are in a fixed order, by kind of entry or by dimension and then alphabetically by title, narrowed only by the filters you set.
• The evidence filter narrows; it never reorders. Its options are “Any peer-reviewed source”, “Reviews or randomised trials only”, “Systematic reviews and meta-analyses only”, and “Only where a review states its own certainty rating”.
• It never picks for you. There is no “choose for me” control; where a step has other options, they are listed beside it and the choice is yours.

The entries are checked by one reviewer, the founder, using risk-based quality assurance: the checks go first to the text where an error could do the most harm, such as safety, dose and referral lines. It is a single reviewer’s check. It is not independent validation.

Every source in an entry’s trail carries its DOI, so any reader can check it too. To see this in practice, open any entry and follow it to the studies behind it. If you find an error, tell us.

In scope: what published research reports people can do for their wellness, cited; and how to evaluate a wellness claim using evidence.

Out of scope: diagnosis, treatment, individual advice, and anything that needs a licensed professional. Each entry names the point at which a question becomes one for a professional. If you are in crisis, call or text 988 in the United States, or 911 for emergencies.