Reads charts, scans and spreadsheets.
Your documents are read the way they were made: tables, scanned pages, sheets, slides and the numbers inside charts.
What gets read, and how.
PDFs and Word files
PDFs are read with their layout and tables, so a figure stays with its row and column. Word files are read with their headings and tables.
Scanned pages
Two AI models from different companies read each scanned page. Lines both agree on are kept. Lines they disagree on are left out and listed, so nothing is guessed. Pages read from images are flagged when you open the original.
Excel, cell by cell
Every cell is read. Sheets of 20 rows or more also get a summary: totals, averages, ranges, and counts and totals by category and by month. All of it is calculated from the cells. Nothing is estimated by a model.
Slides, charts and pictures
PowerPoint text and tables are read, and chart numbers come from the file's own chart data. Charts and pictures elsewhere are read by two models, which must agree on every number.
Web pages and WhatsApp exports
Web pages are read as text. A WhatsApp chat export (.zip) is read with the files shared in the chat.
What it does not do yet. Scanned pages inside files attached to a single question are not read yet. Scans added to the knowledge base are read.
File by file.
| File | What is read |
|---|---|
| Text with its layout and tables. Scanned pages are read by two AI models. | |
| Word | Text, headings and tables. |
| Excel | Every cell. Sheets of 20 rows or more also get a computed summary. |
| PowerPoint | Slide text, tables, and chart numbers from the file's own chart data. |
| Charts and pictures | Read by two AI models, which must agree on every number. |
| Web pages | The page text. |
| WhatsApp chat exports | The .zip export, with the files shared in the chat. |
See it answer from real documents.
The live demo answers from 40 public documents, with 5 free questions. To see it on your own documents, book a demo or see the pricing.