If your budgeting app cannot read a PDF statement, work through three questions in order: is it a text PDF or an image, is a CSV available instead, and does the result reconcile. Most readers solve it at question two. The underlying reason is that a PDF stores a printed page, not a table, so a tool has to rebuild the rows by guessing.
Question 1: is the file text or an image?
- Open the PDF and try to highlight a single word with your cursor.
- If the word highlights on its own, the file has a text layer.
- If the whole page highlights as one block, it is an image, such as a scan or a screenshot saved as PDF.
- If it asks for a code to open, it is protected, and only the owner can open it.
A text layer means a tool has something to work with. An image means it first needs optical character recognition, an extra step that can mix up characters. A protected file cannot be read until you open it yourself.
Question 2: is there a CSV or spreadsheet version?
Look in the place where you downloaded the PDF for other formats for the same period. If one exists, it is nearly always easier for an app to read, since each row and column is written down rather than implied. Say your PDF is a month of chequing activity with 47 lines. The CSV of the same month will have 47 rows plus a header, with each field in its own cell.
| File type | Where the structure lives | Likelihood of a clean import |
|---|---|---|
| CSV | In the file itself: one row per line | High, if headers and dates are tidy |
| Spreadsheet | In cells and columns | High |
| Text PDF | Inferred from word positions | Medium; wraps and page breaks cause errors |
| Scanned PDF | Not present; must be recognised | Lower; characters can be misread |
Question 3: does the result reconcile?
Whatever route you took, test the output. Take a hypothetical statement that opens at C$1,880.00, with credits of C$3,500.00 and debits of C$3,142.60. The closing balance should be 1880 + 3500 - 3142.60 = 2,237.40. Add the rebuilt amounts; if they give 2,237.40 the table is complete. If the sum is out by exactly one amount, you are missing or duplicating that line.
Why do page breaks cause so much damage?
- Each page repeats column headings, which a tool may read as a transaction.
- A running balance carried forward at the bottom of the page may look like an amount.
- A description may start on one page and finish on the next.
- Totals or summaries at the end can be read as a last transaction.
What if I only have the PDF?
Copy the table from the PDF viewer and paste it into a spreadsheet. If every line lands in a single cell, paste into a text editor first to see where line breaks fall, then use Data, Text to Columns with a delimiter or fixed width. Fix the wraps and compare the total. It takes longer than a CSV, but it is manageable for a month at a time.
A tool that shows what it read and lets you fix rows before saving is a better fit than one that imports silently. SpendixAI is built around PDF, CSV and Excel statement import with review before saving, and does not promise that every layout will parse perfectly. It is built around India and INR, so mention it only matters for readers who also keep Indian accounts, not Canadian ones.
The short version
- Highlight a word to test for text.
- Ask for a CSV first.
- Join wrapped lines.
- Reconcile with the closing balance.
- Keep the original.
Useful ideas start with a clearer picture of your own spending.
Explore your reports


