Solution for Research and analysis
Forty papers, one evidence table.
The answer is somewhere in the methods, the results and occasionally only the supplement of every one of them, reported differently each time and sometimes not at all. Ragextract reads the corpus once and returns the column, including the rows where it was never stated.
Free credits on sign-up. No card, no subscription.
| Row | Documents | Sample size | Methodology | Risk of bias | Funding conflict | Year |
|---|---|---|---|---|---|---|
| Cohort study — Rennick.pdf | 1,284 | Prospective cohort, 18-month follow-up | Attrition 22%, not adjusted for | Yes — industry-funded | 2024 | |
| Meta-analysis — Okonkwo.pdf | 9,402 | Random-effects meta-analysis of 17 studies | Funnel plot asymmetry noted by authors | No | 2025 | |
| RCT — Vasquez.pdf | 312 | Double-blind randomised controlled trial | Low — pre-registered, intention to treat | No | 2023 | |
| Preprint — Halloran.pdf | 88 | Single-site crossover, not peer reviewed | Not stated | Not stated | 2026 |
A figure you can put in the write-up.
Research output lives or dies on attribution. A sample size is worth having only if the page it was reported on travels with it, and a summary that cannot be traced back is a summary somebody else has to redo.
Every cell carries its page and a confidence score, so the checking goes to the answers that look thin rather than back through the corpus. What the table changes is the ratio of reading to writing, not whether the reading happens.
| Document | Sample size | Funding conflict |
|---|---|---|
| Cohort study, 2024 | 1,284 | Yes — industry-funded |
| Meta-analysis, 2025 | 17 studies, 9,402 participants | No |
A row is one subject, one period.
This is the decision that makes a corpus into a series. One research group across four publications is four rows, not one, and a year column is what lets the table be read down as well as across — which is where the interesting answer usually is.
Bundling is the exception rather than the rule here. A paper and its supplementary material are one study and belong together; two papers from the same group are two studies and do not. Bundle the first, never the second.
Set the output type to match what you will do with the answer. Sample size is a number so it can be sorted; reported outcomes are a list because there are usually several; “not reported” is a legitimate text answer and a much more useful one than a blank.
Two of the types answer with a picture rather than a sentence, which is what the literature case has always wanted. The Academia preset asks for every figure presenting a study’s results, with its axis labels and legend, and puts them in the row — so a methodological comparison across forty papers can be looked at rather than described.
Where it fits.
Two very different corpora, one job: find the handful of fields that make the members comparable, and put them side by side.
| The work | What a row is |
|---|---|
| Literature and evidence review | A published paper |
| Systematic review screening | A candidate study |
| Competitive and market scanning | A filing, report or release |
| Longitudinal tracking | One subject, one period |
These templates ship in the application, and they are the ones not yet worked through under use cases — so start from columns, which covers the output types and how to write a prompt that holds up, and take clinical guidelines as the closest worked setup for the shape: a corpus of published documents, read for what makes them comparable.
Reading is charged once.
A corpus is billed for the pages it reads, not for the questions asked of them. The question a reviewer raises in month three, asked of papers ingested in week one, costs only its own column — which is the difference between a corpus you maintain and one you rebuild.
| A 40-paper literature set | Volume | Cost |
|---|---|---|
| 40 papers, 25 pages each — read once | 1,000 pages | 1,000 credits |
| 3 typed questions asked of all 40 | 120 cells | 480 credits |
| Reported outcomes, as a list column | 40 cells | 480 credits |
Nothing on this page is investment or research advice, and the presets are not methodological definitions.
Start with the pile you already have.
Signing up comes with free credits. No card, no subscription.