Skip to main content
Point Anara at a folder of twenty, forty, or a hundred papers and run the same structured extraction across all of them at once. Anara returns a combined table, one row per paper, populated from the sources you uploaded. For the moment you have finalized a template and just need it applied at scale.

​
1. Describe the task

You already know the template. Title, authors, year, sample, intervention, outcome measures, effect size, limitations. You have run it one paper at a time for the first ten and it works. You have sixty papers left and no appetite to open sixty more chats. Anara runs the same extraction across every paper in a folder and returns a single combined table. You define the template once, point Anara at the folder, and get back rows populated from each source with citations. This is a Deep Search workflow, so it requires the Max plan. Here a PhD candidate doing a systematic review on compassion fatigue interventions in healthcare has a folder of sixty included studies and a ten-field extraction schema already validated on her first batch. 
I need to run my extraction template across every paper in my "Compassion Fatigue SR - Included" folder. It has 60 papers. Use Deep Search across the folder so this runs in parallel.

For each paper, extract these fields:
- Title
- First author and year
- Country and healthcare setting (ICU, primary care, oncology, etc.)
- Study design (RCT, quasi-experimental, qualitative)
- Sample size and population
- Intervention description (what was done, duration, dose)
- Primary outcome measure and scale used
- Effect size with confidence interval if reported
- Key limitations noted by the authors
- Risk of bias rating (using Cochrane RoB 2 if reported, otherwise leave blank)

Return a single combined table with one row per paper. Where a field is not reported in the paper, leave the cell blank and do not infer. At the end, flag any paper where three or more fields came back blank so I can review whether it belongs in the review at all.

​
2. Give Anara context

Required context
  • A folder in your library containing the papers you want to extract from.
  • The full extraction template, stated in the prompt or pasted from a completed example row.
Optional context
  • One or two already-completed rows as examples. Anara matches your phrasing and level of detail.
  • Your inclusion criteria document. If any papers look like they do not fit, Anara flags them with a note.

​
3. What Anara creates

A single combined extraction table with one row per paper in the folder. Each cell is populated from the paper or left blank with an explicit “not reported” note. The output is the raw material for your systematic review results table, ready to paste into your analysis software or dissertation appendix. Papers that triggered the blank-field threshold get flagged at the bottom so you can triage inclusion decisions before the analysis phase.

​
4. Follow-up prompts

​
Dig into one paper that looks off

When a row comes back with unexpected values or too many blanks. 
Row 23 (the Johnson et al. 2019 paper) has the effect size missing and an unusual sample description. Open that paper specifically and tell me what the paper reports about the primary outcome, why you left effect size blank, and whether this paper should be excluded on methodological grounds.

​
Turn the table into the results section

When the extraction is clean and you are ready to draft. 
Based on the combined extraction table, draft the Results section for my systematic review. Structure it as: sample characteristics across studies, intervention types, primary outcome effects, and risk of bias summary. Use PRISMA-style reporting. Cite every claim to the row it came from.

​
Correct a field across the whole table

When you realize one of the template fields was interpreted differently than you intended. 
For the "healthcare setting" column, I want you to re-extract across all 60 papers using a stricter definition: only use ICU, ED, oncology ward, or primary care as the four allowed values. Anything else, put "other" and note what the paper actually described.

​
5. Tricks, tips, and troubleshooting

​
How you word your prompt shapes what you get

State the template fields explicitly and name the expected format for each. “Effect size with confidence interval if reported” produces structured output; “effect size” alone produces inconsistent formatting across rows. Tell Anara what to do with missing data up front (“leave blank, do not infer”), otherwise it may fill gaps with plausible-sounding values that are not in the paper. For folders over thirty papers, consider splitting into sub-folders of ten to fifteen and running separately. Deep Search parallelizes across a folder but does not guarantee full coverage at very large sizes.

​
Check the output against your own understanding

Spot-check five rows against the source papers before trusting the full table. Pick the two papers you know best, one paper where the extraction looks suspiciously clean, and two papers chosen at random. If those five match what the papers actually report, the remaining rows are likely reliable. If any of the five have a cell that does not match the paper, run that row again on its own and ask Anara to quote the source passage for each field.

​
What to do with the output next

Paste the combined table into your review’s analysis software or dissertation document. For Cochrane-style reviews, the flagged “three or more blanks” papers need a manual decision before the analysis phase. For narrative reviews, the combined table becomes the backbone of your Results section, which the follow-up prompt can draft directly. Save the extraction conversation itself as a note so you can re-run the same template across additional papers added to the folder later.