Feature

Chat with PDF: ask questions to research papers and your own files

Upload a PDF or another document to the library, select it as a source and ask: what is the method, what are the results, where are the limitations. The answer relies on the file text and shows where each claim comes from.

scid.ai library with uploaded research paper PDFs
The libraryUploaded PDFs next to search, filters and the Upload button.

Upload a PDF to the library

The file appears in the list with its format and size. From here it can be attached as a source to any research session, alone or together with papers found in search.

Ask a question and check where every claim comes from

The answer arrives with a source reference on every claim. Clicking the reference opens the passage it was taken from, so a claim can be verified before it goes into a review.

scid.ai chat: an answer with source references and the sources panel on the right
The answerThe answer with source references on the left, the sources panel with cards on the right (Russian interface shown).

How chat with PDF works in scid.ai

The file goes into the research library and becomes a source in the same way as a paper found in search. You select one or more documents and the chat answers only from them: no guessing and no unrelated material mixed in.

Every claim in the answer is tied to a passage of the file. You can open it, check the context and only then use it in a review, a table or a document.

  1. Upload the document. A paper PDF, a DOCX report or a spreadsheet: the library accepts working formats and keeps the file next to the research.
  2. Select it as a source. Tick the file in the sources panel and add found papers or other files if you need a comparison.
  3. Ask a question. "Which method was used", "which limitations do the authors admit", "compare the conclusions of two reports". The answer comes with passage references.
  4. Keep going. Turn the answer into a table, include it in a literature review or save it to a document with a bibliography.

What it is for

  • Understand a long paper quickly: method, sample, results, limitations, what remains open.
  • Extract specific data from a PDF: experiment parameters, numeric results, term definitions.
  • Compare two or more documents: where the authors agree, where they contradict each other, how the methods differ.
  • Check your own draft against sources: does the paper support the claim you cite.
  • Work with tables: ask a question to CSV or Excel rows and get a summary of the data.

How this differs from uploading a PDF to a general chatbot

A general chatbot answers from the file but does not keep it in a research context or link it to the literature. In scid.ai the file lives next to the found papers, the topic graph and the review: you can ask how the conclusions of an uploaded report relate to publications on the topic and get an answer with sources on both sides.

Full-text access to third-party papers depends on the licence: scid.ai does not bypass paywalls. You upload your own PDF; for found papers, open versions and metadata are used.

Frequently asked questions

Which formats are supported?

PDF, TXT, Markdown, DOC and DOCX, RTF, CSV and TSV, JSON, XML, HTML, TeX and Excel workbooks. Scans without a text layer need OCR first.

Can I ask several PDFs at once?

Yes. Select several files as sources and the chat answers across all of them, noting which document each claim comes from.

Do I need a paid plan?

Chat with files is available on the free plan within the starting credits; sign-up needs no card.

Can the answer be verified?

Yes, every claim leads to a passage of the file or to the paper page. scid.ai does not guarantee that a generated statement is always correct, so the source is always next to it.

Ask your PDF right now

Upload a paper or a report and ask the first question. Free, no card.

Upload a file