Literature review matrix
Build a synthesis matrix across 5–30 uploaded papers: eight fixed evidence columns, verbatim quotes with page and character locators, and reference-only source packages for Writer Agent and Integrity Review.
Updated 2026-07-22
How the paper library, evidence locators, source packages, and reference workflows work — and where ChatPDF ends and the matrix begins.
Yes. The Literature Review Matrix requires a WriteGo account so the paper library, evidence matrix, and references share one authorization boundary. It is currently free: uploading papers, building the matrix, cross-paper search, and reference export charge no credits.
No, never. Uploading papers, building the evidence matrix, and managing references do not call the AI Detector or create detection records. The matrix also has no AI chat: every contentful cell holds claims backed by verbatim quotes captured from your uploaded papers, under visible review states rather than model-generated summaries or scores.
A source package is immutable, versioned, and reference-only. It carries paper fingerprints, page numbers, and character offsets — never the full text of your papers. Writer Agent and Integrity Review can resolve exactly the evidence entries you selected, and evidence-level return links lead back to the matrix entry each claim came from.
ChatPDF is a single-paper reader with AI question answering; the Literature Review Matrix synthesizes evidence across 5–30 papers and has no AI chat. Each tool keeps its own paper library, isolated on the backend, so neither can see the other's files. Read one paper closely in ChatPDF; compare many papers in the matrix.
Compare 5–30 uploaded papers across research question, sample, method, result, effect, limitation, conflict of interest, and DOI. Candidate rows auto-seed from extracted paper metadata checkpoints, so the matrix starts from what the papers state rather than an empty grid.
Every contentful cell holds claims backed by verbatim captured quotes with paper-version, page, and character-offset locators. Unreviewed, not-located, needs-review, and conflicting evidence stay visibly distinct, and the matrix never adds a black-box quality score. Export to Markdown or formula-safe CSV.
Upload PDFs up to 10 MB. Server-side text extraction runs for every paper, with server OCR for image-dominant pages, and each paper shows its processing status with retry and delete. The library is fully separate from ChatPDF's — backend product isolation means neither tool sees the other's files.
Cross-paper search runs on the server with deterministic ranking over the papers you uploaded. Results stage as pending matrix evidence, so a useful passage becomes a locatable claim instead of a copy-pasted fragment.
Hand selected evidence to Writer Agent or Integrity Review as an immutable, versioned package containing paper fingerprints, page numbers, and character offsets — never full text. Evidence-level return links reopen the exact matrix entry a claim came from.
Resolve metadata from a DOI, arXiv ID, PubMed ID, or URL, or import Zotero JSON, BibTeX, and RIS. Export APA 7, MLA 9, Chicago, BibTeX, RIS, or Markdown. Every field keeps its provenance, so you can see whether a value was resolved, imported, or edited.
The Literature Review Matrix never runs AI detection and has no chat assistant. It is an evidence-organization surface: what enters the matrix is what your papers verifiably say, kept under review states you control.
Currently yes. It requires sign-in, but uploading papers, building the matrix, cross-paper search, and reference export charge no credits today.
Eight fixed columns: research question, sample, method, result, effect, limitation, conflict of interest, and DOI. The column set is fixed so matrices stay comparable across projects.
No. The Literature Review Matrix keeps its own paper library, isolated from ChatPDF's on the backend. Upload the papers you want to synthesize directly into the matrix library.
Yes. The matrix exports to Markdown and formula-safe CSV. Per-paper references export to APA 7, MLA 9, Chicago, BibTeX, RIS, and Markdown, with per-field provenance preserved.