How to Compare Sources Side by Side With Afforai
Afforai compares documents by letting you put several PDFs in one library and asking a question that spans them, such as which of five papers reports the strongest effect size. Every claim comes back with a citation you can click to see the source sentence. A five-paper comparison takes about ten minutes, most of it reading the citations.
The comparison mode is where Afforai beats a single-document chat tool. Most PDF assistants answer inside one file at a time; a literature review needs the cross-file view.
What does the free tier include?
Free accounts get a limited page allowance per month, roughly enough for a few hundred pages of uploads, plus a cap on the number of questions. Paid tiers start around $10-$20 a month depending on the page quota, with an annual discount.
For a single literature review, free usually covers it. For a thesis with 80 papers, you will hit the page cap inside the first month.
| Free | Paid | |
|---|---|---|
| Pages/month | ~200 | 1,000+ |
| Documents per library | Limited | Higher |
| Export | Restricted | Available |
How do you set up a comparison library?
Create a library per research question, not per project. Mixing 40 papers from three different subtopics produces muddy answers because retrieval pulls the wrong ones.
Steps:
- Create a new library named after the question, e.g.
sleep-and-memory-consolidation - Upload the PDFs, waiting for each to finish parsing before adding the next
- Tag or group papers by finding type if the tool supports it — RCTs versus observational studies behave differently
- Ask the contrast question with the comparison framing explicit
Step 2 matters more than it looks. Uploading 20 files at once often leaves three silently unparsed, and you will not notice until an answer is missing a key paper.
How should the comparison prompt be phrased?
Name the papers and the dimension. “Compare these” returns a summary of each. “Compare these on sample size and reported effect size, and flag disagreements” returns something usable.
A prompt that works:
Across these five papers, extract sample size, population, intervention duration and the main reported outcome. Then state where the conclusions conflict and quote the sentence that supports each position.
Asking for the quote changes the output from paraphrase to evidence. It also makes verification fast, because you can jump to the sentence rather than re-reading the paper.
What should you verify before citing anything?
The citation, every time. Afforai highlights the source sentence, and the answer is only as good as whether that sentence says what the summary claims.
Three failure modes to watch for:
- Table extraction errors — numbers in a results table often come through shifted by a row
- Two-column PDFs — the parser interleaves columns and produces sentences from both
- Reference sections — a citation to a paper cited inside your paper can be mistaken for that paper’s own claim
Downloading and opening the PDF for any number you plan to put in your own write-up is not optional. It is the job.
Does it handle scanned PDFs?
Partially. Text-based PDFs parse cleanly. Scanned images need OCR, and quality drops sharply on old journal scans with dense two-column layouts.
If your sources are mostly pre-2000 scans, test with two documents before uploading the whole set. A garbled parse produces confident nonsense, which is worse than no answer.
How does it compare to a general chat tool?
A general chatbot will summarize a PDF you paste. Afforai organizes files into libraries, keeps retrieval scoped, and enforces citations in its answers by default.
For one paper, the chat tool is faster. For 15 papers and a question that spans them, the library structure is the difference between an answer and a guess.
Where does Afforai fall short?
It cannot read data behind a paywall, and it cannot judge methodology quality. Two studies with the same effect size but wildly different rigor come back treated as equally credible unless you ask about design explicitly.
Exports are limited on free plans, which matters if you need your notes in a specific format for a review paper.
Who is this for?
Graduate students and researchers doing systematic comparisons across 10-50 papers. Analysts comparing competitor reports or policy documents from several organisations.
Not for anyone who needs to cite without opening the source, and not for teams needing shared editorial control over a research library.
Quick Summary
- Afforai compares across multiple documents in one library and returns clickable citations.
- Free tier is roughly 200 pages a month; paid starts near $10-$20.
- One library per research question keeps retrieval focused.
- Table numbers and two-column PDFs are the main parsing failure points.
- Always open the cited PDF before using a number in your own work.
