Document Analysis Troubleshooting
Diagnostic procedures for document upload failures, unreadable files, incorrect document-based outputs, table extraction errors, context loss on large files, and processing delays.
Step-by-step fixes for problems with uploading, reading, and analyzing documents. Find your symptom, follow the steps, get back to work.
Quick Diagnosis
Start here. Find your symptom, then jump to the detailed fix.
| Symptom | Most likely cause | First step | Section |
|---|---|---|---|
| File won't upload | Size, format, or file corruption | Check size (≤20 MB) and format | Upload Failure |
| "Can't read file" error | Legacy format, password protection, or image-only PDF | Convert to PDF or DOCX | Unreadable File |
| Answer doesn't match document | Underspecified prompt or hallucination | Ask IAN to quote its source | Incorrect Document Output |
| Table data is wrong or jumbled | PDF table stored as visual layout | Describe columns explicitly, or paste data as text | Table Extraction Errors |
| Later answers become vague on long docs | Document exceeds context window | Split into sections and process separately | Context Loss on Large Documents |
| Processing spinner runs past 60 seconds | Large or complex file | Refresh, try a smaller file or different format | Processing Delay |
| Uploaded the wrong file | — | Start a new chat | Wrong File Uploaded |
1. Upload Failure
What you see: You click [Attach], select a file, and get an error — or nothing happens.
Pre-check
Run through this checklist first. Most upload failures are caused by one of these four issues.
| Check | What to look for | How to fix |
|---|---|---|
| File size | Must be under 20 MB (under 10 MB is recommended) | Split the file into smaller parts or compress it |
| File format | Must be PDF, DOCX, TXT, CSV, or XLSX | Save as one of these formats |
| File name | Special characters or very long name | Rename: remove symbols, keep it short |
| File integrity | Can you open the file normally on your computer? | Re-download or re-save the file |
Resolution
Try each step in order. Stop when the issue is resolved.
| Step | Action |
|---|---|
| 1 | Check the file size (right-click → Properties on Windows, Get Info on Mac) |
| 2 | If over 20 MB, split the document or compress it |
| 3 | If the format isn't supported, save as PDF or DOCX |
| 4 | Re-save the file (open → File → Save As → save a fresh copy) |
| 5 | Try a different browser |
| 6 | Try drag-and-drop instead of the [Attach] button (or vice versa) |
If the file is too large to split
- Copy the most relevant sections into a new document.
- Upload that smaller document.
- Ask your question about those specific sections.
2. Unreadable File
What you see: The file uploads successfully, but IAN says it can't read it — or gives an answer that clearly has nothing to do with your document.
Common causes
| File type | Why it fails | How to fix |
|---|---|---|
.doc files (pre-2007 Word) |
Outdated format not supported | Open in Word → Save As .docx |
| Password-protected PDFs | IAN cannot decrypt protected files | Remove the password, re-save, then upload |
| Scanned documents (no OCR) | Contains images of text, not actual text | Run OCR first (Adobe Acrobat or any OCR tool), then upload |
Proprietary formats (.pages, .numbers, .odt) |
Not supported | Export to PDF or DOCX |
| Image-only PDFs (no text layer) | Same as scanned documents | Run OCR on the PDF |
Resolution
| Step | Action |
|---|---|
| 1 | Open the file on your computer and confirm you can read it |
| 2 | Try selecting and copying text from the file — if you can't select text, it's image-based and needs OCR |
| 3 | Save a fresh copy in PDF or DOCX format |
| 4 | Upload the fresh copy |
Quick check for PDFs: Open the file and press
Ctrl+A(Select All). If text highlights, the PDF is searchable and IAN can read it. If nothing highlights, it's image-based and needs OCR processing first.
3. Incorrect Document Output
What you see: IAN gave an answer about your document, but it doesn't match what you actually see in the file.
Resolution
| Step | Action | What to look for |
|---|---|---|
| 1 | Ask for a direct quote: "Quote the exact text where you found that answer" |
Verify whether IAN is citing real content |
| 2 | Open your file and find the quoted section | Does the quote actually exist in the document? |
| 3a | If the quote is accurate but the interpretation is wrong: rephrase with more context (see examples below) | — |
| 3b | If the quote is wrong or fabricated: the model hallucinated — try a more specific question, or upload just the relevant section | — |
Rephrasing examples:
| Situation | What to type |
|---|---|
| IAN misunderstood the meaning | "Based on the sentence '[exact quote]', does this mean X or Y?" |
| IAN used the wrong context | "I'm a claims adjuster reviewing this. When it says [term], in our context that means…" |
| IAN looked at the wrong section | "In the section titled 'Coverage Exclusions', what does it say about…" |
Common causes of wrong answers
| Cause | What happens | How to prevent |
|---|---|---|
| Vague question | IAN guesses instead of finding specific text | Be specific: "What does Section 3.2 say about X?" |
| Ambiguous document | Multiple sections could answer the question | Point IAN to the right place: "In the section titled 'Coverage'…" |
| Very long document | Earlier content falls outside the context window | Split into sections and ask about each separately |
| Complex formatting | Tables, multi-column layouts, or headers get misread | Describe the structure: "This is a two-column table with…" |
Important: Always verify document-based answers for claims, underwriting, compliance, or legal work. IAN is a research assistant, not a decision-maker.
4. Table Extraction Errors
What you see: You ask about numbers in a table, and IAN returns wrong values or mixes up rows and columns.
Why it happens: Tables in PDFs are often stored as visual elements, not structured data. IAN has to infer where columns and rows are, and it doesn't always get it right.
Resolution
| Step | Action |
|---|---|
| 1 | Describe the table structure explicitly: "This is a table with columns: Date, Region, Revenue, Expenses. Row 1 is headers. Data starts at row 2." |
| 2 | Ask about specific cells instead of the whole table: "What is the Revenue value for Q2 in the East region?" rather than "What are the numbers?" |
| 3 | If that doesn't work, extract the table manually: open the file, copy the table data, paste it into a new chat message, and ask IAN about the pasted text |
Spreadsheets vs. PDF tables
If you have the same data in both a spreadsheet and a PDF, upload the spreadsheet (CSV or XLSX). Spreadsheets preserve column and row structure, so IAN reads them far more reliably than PDF tables.
5. Context Loss on Large Documents
What you see: You uploaded a long document (50–100+ pages). IAN answers the first few questions well, but later answers become vague, repetitive, or miss information.
Why it happens: Every model has a context window — the maximum amount of text it can process at once. Documents that exceed this limit lose earlier content from the model's working memory.
Resolution
| Step | Action |
|---|---|
| 1 | Split the document into 3–5 sections by topic or chapter |
| 2 | Upload one section per chat session |
| 3 | Ask your questions about each section and note the answers |
| 4 | Start a final chat, paste in your findings, and ask: "Synthesize these into one overview" |
For repeated use of large documents
If your team needs to query the same large documents regularly, use Knowledge Bases. Knowledge Bases automatically break documents into searchable chunks, avoiding context window limitations.
6. Processing Delay
What you see: You uploaded a file and IAN has been "processing" for over 60 seconds.
Expected processing times
| File size | Typical processing time |
|---|---|
| Under 5 MB | 5–8 seconds |
| 5–10 MB | 8–15 seconds |
| 10–20 MB | 15–30 seconds |
Resolution (if past 60 seconds)
| Step | Action |
|---|---|
| 1 | Refresh the page |
| 2 | Start a new chat and re-upload the file |
| 3 | Try a smaller file or a different format (PDF typically processes fastest) |
If the problem persists, contact support with the file size, format, and how long you waited. Email: gss.sr@xceedance.com.
7. Wrong File Uploaded
What to do: Start a new chat and upload the correct file.
You cannot remove an attached file from an existing conversation. The old chat remains in your History — delete it if you no longer need it.
Reference: Multiple File Uploads
You can attach multiple files in a single chat, up to 50 MB total across all files.
To compare documents: Upload both files, then ask:
"How do these two documents differ on [topic]?"
Escalation
If none of the above resolves your issue:
- Note the file name, size, and format.
- Note the error message (exact text or screenshot).
- Note what you asked and what IAN responded.
- Submit via Support & Escalation. Email: gss.sr@xceedance.com.
Related Pages
| Page | Purpose |
|---|---|
| Document Analysis | Full feature reference for document upload and analysis |
| Knowledge Bases | Persistent, chunked document storage for repeated queries |
| Chat Interface Troubleshooting | Fixes for chat-level issues (not document-specific) |
| Support & Escalation | How to reach a human for unresolved issues |