A quick reference for reading and evaluating research literature. Written for master's students starting a thesis or research project, and for anyone newer to the process.
Most of what makes a claim trustworthy or shaky lives in a handful of recurring statistics and a few recurring pitfalls. Tap a card to see what each one really means.
What a confidence interval is telling you
Same made-up result (an odds ratio of 0.7), with more and more participants. Drag the slider.
95% CI: 0.47 to 1.05. The interval still crosses 1.0, so the data can't rule out no effect.
Every paper you add gets four fields, plus the exact passage the summary was drawn from. Treat the summary as a fast first pass, not a substitute for the source:
The paper's central finding, in one or two sentences. Check it against the source excerpt. A good summary is specific (what changed, by how much, compared to what), not a vague restatement of the topic.
Who or what was actually studied. This matters as much as the finding: a result in mice, or in a narrow patient subgroup, doesn't automatically apply to the people you care about.
The study design in brief. Check it against the Study types guide: the design tells you how much weight the finding should carry before you've read anything else.
The specific numbers behind the claim: effect sizes, confidence intervals, p-values and sample sizes. This is the field most worth double-checking against the abstract yourself, since it is the easiest place for a summary to lose nuance, like whether a CI crossed 1.0.
Every paper in your library is labeled with how much of it this app could read, legally:
Full text (OA)
The paper is open access, and the summary came from the article itself, not just the abstract.
Uploaded full text
You uploaded the PDF or document, so the summary and your questions can use the whole thing.
Abstract only
The paper is behind a paywall this app doesn't bypass, so the summary is based on the published abstract, which can leave out methods. Upload the PDF if you have access.
Metadata only
No summary was possible. You have the citation and a link to the source.