Document Reader
Ask a question about document text and review the answer beside exact source excerpts.
Fill in the fields, then submit when you are ready.
Result
Your result will appear here after you submit and complete verification.
How to use Document Reader
- Choose a PDF, DOCX or UTF-8 text file under Read a document on this device, then select Read file locally. Alternatively, paste Document text and use Pasted text as Source format.
- Review the text before submitting. The local reader sets Source format and adds PDF page or DOCX paragraph markers; these markers count toward the 18,000-character source limit.
- Enter Question about the source for a narrow answer, or leave it blank for a bounded summary. A PDF needs a readable text layer; this local reader does not OCR scanned pages.
- Select Read document and complete browser verification when prompted. The reviewed text is sent for processing; the original local file is not uploaded by the reader.
- Compare the answer with each cited excerpt and the original document. Source locations identify submitted text, not a guarantee that extraction order, interpretation or the underlying reporting is correct.
Example: Document Reader
Ask a narrow question about a short workshop note.
Options
- Local file reading
- PDF page markers refer to actual PDF pages. DOCX markers count body paragraphs, including empty ones; they are not Word page numbers.
- Question
- Ask about one fact or passage. Leave the question blank for a bounded summary rather than requesting unrelated research.
Supported inputs and limits
- Document text allows 18,000 characters, including source markers;
- Question about the source allows 1,000.
- Local files must be nonempty and no larger than 10 MiB.
- PDF input is limited to 20 pages with a text layer.
- DOCX input allows 200 body paragraphs, including empty ones.
More details
TXT must be readable UTF-8 without binary control characters.
DOCX tables, drawings, equations, tracked changes and external relationships are refused.
Headers and footers are omitted, and DOCX markers are paragraph numbers rather than Word page numbers.
PDF reading order and spacing can differ from the original page.
Large source maps can require a shorter excerpt even within the character limit.
The original file stays on your device; the text you review and submit is sent to the configured AI service.
Source excerpts establish where words occur in that submitted text, not independent accuracy or a complete document reading.
Check the extraction and answer before relying on them.
Where your input is processed
Your input is uploaded to this site and sent to a configured AI provider. If a provider fails, the request may be sent to the next eligible provider. Check its result before use. Private inputs and results expire within one hour and are excluded from site backups; temporary files are removed by scheduled cleanup. Provider retention rules may differ. Avoid submitting passwords or highly sensitive information.
Check the extraction before the answer
Compare the extracted paragraphs with the original document first. Reading order can change in columns, footnotes may be absent and line breaks may differ. Keep only the passages needed for your question. After processing, read each quote in context, especially exceptions, dates and conditions nearby. Resolve unclear wording in the original document instead of treating a source label as proof of the whole answer.
Questions about Document Reader
Are citations taken from the actual file?
They point to the text you reviewed and submitted. PDF labels retain page numbers; DOCX labels retain paragraph numbers. A quote confirms where words occur, not that the generated interpretation is correct.
Can it answer a question the document does not cover?
It should report missing evidence and ask for detail. Check the excerpts yourself; a fluent answer can still overstate the source.
Can I upload a scanned PDF?
The local reader needs a text layer. Use Image and Handwriting OCR for individual scanned pages, review that transcription, then paste the relevant text here.