Library

Your Library (sidebar → Library) holds every paper you’ve touched: papers the assistant cited in your chats, papers you added yourself, and papers imported from Zotero. It’s private to you and spans all your projects.

It has four tabs: Library, Search, Radar and Sync. This page covers the first two. See Radar and Zotero for the others.

The paper list

Papers are grouped by research question by default; switch to grouping by project, or no grouping, with the Group menu. The Filter by status menu narrows the list to Included, Maybe, Excluded, Unscreened, or papers not attached to any question.

The search box searches your whole library, not just the rows on screen.

Each row shows a verification dot, the paper (title, authors, venue, DOI), year, project, tags, status, relevance and when it was added. The authors read as in a reference list: up to ten names, then et al. when the paper has more. Papers added since your last visit carry a New pill, and the sidebar’s Library entry shows how many are waiting.

Hover the Verified column header for what the colours mean; the full explanation is in Papers and verification.

From your chats

Papers the assistant cited in a chat land automatically in a group at the top called From your chats. Click Turn into a research question on that group to create a question and attach all of them to it, ready for triage.

Research questions

A research question is the unit of organisation: a statement of what you’re trying to find out, optionally tied to a project. Papers are attached to questions, triaged against them, watched by Radar for new publications, and exported per question.

Create one with + Add → Research question, or ask the assistant: “Track a research question about X”. The assistant will propose one and only create it when you agree.

Adding papers

+ Add → Paper opens a dialog where you choose a project and (optionally) a question, then supply the paper as a DOI, a URL, or an upload (PDF, Word, text or Markdown). An upload without recognisable metadata still becomes a Library item — this is the deliberate way to bring in an unpublished manuscript, a report or a thesis.

Files you attach in a project chat are not added to the Library automatically unless they are recognised as a paper: a PDF whose DOI, arXiv id or title matches a registry record. Pasted text, notes, spreadsheets and unrecognised PDFs stay project documents (the assistant can still read them); add one to the Library deliberately with + Add → Paper → Upload if it belongs there.

For an upload you also pick what kind of work it is — Manuscript (the default, for unpublished work such as a draft or a student paper), Preprint or working paper, Report, Thesis or Other document — and can enter the title (prefilled from the filename), authors (separate several with ;), year and the institution, repository or university. The file becomes a Library item with those details and shows its type where recognised papers show a verification dot; exports use the matching type (BibTeX @unpublished, @techreport, @phdthesis; Zotero Manuscript, Report, Thesis). If the file is later recognised as a published paper — a PDF whose DOI, arXiv id or title matches a registry record — its registry record replaces the entered details.

To bring in a reference list, open a research question and choose Import references: pick a BibTeX (.bib), RIS (.ris) or CSV file — the formats reference managers and databases export — with up to 50 records. Each record with a DOI, PMID or arXiv id is added to the question; one without is looked up in CrossRef by author, year and title and added only when the title matches. The file’s titles and authors are replaced by the registry record once the paper is verified. A summary lists what was added, what was already in the question, duplicates in the file, and every record that could not be read or identified, with its line.

Papers also arrive from:

• the + Add button on a paper card in a chat;
• + Library in a chat’s Citations panel;
• the Search tab (see below);
• Radar and Zotero imports.

Triage

Every paper attached to a question has a status: Unscreened at first, then Included, Maybe or Excluded. Change it from the pill on the row, in the paper drawer, or on the paper’s page.

Excluding a paper asks for a reason, chosen from a fixed list — wrong intervention, wrong population, wrong outcome, wrong study design, duplicate, or not retrievable — so you have a clean audit trail for a PRISMA-style review.

Nothing is ever included automatically. Papers from chats, Radar and imports all arrive unscreened.

To triage several at once, tick the rows and use the bar that appears: Include, Maybe, Exclude… (within a question group), or + tag selected anywhere.

Screening in stages, with one or two screeners

For a systematic review, open the research question and use its Screening section. Papers are screened on their title and abstract first; a paper you include or mark maybe there moves on to full text, and the full-text decision becomes its status. The section counts, for each stage, the papers still to screen, those in conflict, and those decided, with the exclusion reasons.

Next to screen shows one paper at a time with its abstract: choose Include, Maybe, or Exclude… with a reason from the list above and an optional note.

With Two screeners, each paper needs two independent votes that agree. You don’t see the other screener’s vote until you have cast your own, and once you’ve seen it your vote can’t be changed. When the two disagree the paper appears under Conflicts to resolve with both votes, and anyone who can edit the project settles it; the resolution records who decided and when. On a two-screener question the triage pill and the bulk bar are refused: the votes decide. The number of screeners can’t be changed once anyone has voted.

The PRISMA 2020 flow

Below Screening, the research question shows its PRISMA 2020 flow diagram, counted from the screening as it stands: records identified (by where each paper came from), records screened and excluded on title and abstract, reports sought, not retrieved and assessed at full text, the exclusions per reason, and the reports included. Where CoChat cannot know a number — duplicates merged when a paper was added, papers removed from the question, which reports belong to the same study — the box says not recorded instead of guessing. Download SVG saves the diagram for a manuscript; the PRISMA CSV on the project’s Papers tab lists every paper with the box it counts in.

The review queue

On the project’s Sources tab, Review queue on a question lists the papers that still need your attention: papers you haven’t marked reviewed, checkup and stats findings you haven’t confirmed or dismissed, claims whose verdict is misattributed, disputed, partly supported or unverifiable, and retracted papers or papers with an expression of concern. The queue is yours alone: another member’s reviews don’t clear it. Retracted papers come first, then papers with recomputed findings (arithmetic, statistical, unit), then flagged verdicts, then the rest. Excluded papers are left out.

Each row says how many of the paper’s references have been checked, for example 3 of 12 references checked. Reference checks run from a shared daily pool, so most papers have only some of their references checked. Click a row to open the paper on the part that needs you, and review it there. Marking a paper reviewed clears its verdicts and retraction from your queue until one of them changes. When one comes back, your status on the paper reads Reviewed · changes since; to clear it, set your status back to Read and then to Reviewed again.

The Search tab

Search the academic databases directly from the Library, without a chat. Each result can be added to your Library. Under a search you can also choose Get alerts for this search → Watch on a question, which makes Radar re-run it every week.

Tags and notes

Tags are yours — add them from the row’s detail or the paper page, with suggestions from tags you’ve used before. Zotero tags import as tags here too.

Each paper’s Notes tab has two fields, saved when you leave them. Your notes are only yours: nobody else sees them, project members included. Shared with your project appears when the paper is on a project question, and everyone on the project can read it; members with write access can edit it. Zotero note sync mirrors the shared field, as before. Notes written before the split were stored in the shared field and stay there; on a paper outside any project they show as Earlier notes, still visible only to you.

The status bar under the title starts with Your status: click it to move the paper through Unread, Reading, Read and Reviewed. Reviewed means you have gone through it (and have read it). Your status is yours alone.

Export

The Export menu in the header downloads your whole library as BibTeX, RIS (which imports into Zotero) or CSV.

For a single question, export from Project → Knowledge → Sources. BibTeX and RIS exports there include only papers that have been verified as real; CSV includes everything, with a verification column.

The paper drawer

Click a row to open a drawer with the paper’s details in seven tabs: Overview, Notes, Claims, Citations, Question fit, PDF and Metadata. Each is explained in Papers and verification. The drawer has its own URL, so you can link to it.

Limits

There’s no cap on how many papers your library can hold. The list shows up to 500 papers at a time — use search to find anything beyond that. Daily allowances apply to citation and claim verification and to uploads; see Plans & limits.