How to tell if a PDF is scanned before converting it to Excel
You can usually tell whether a PDF is ready for a spreadsheet by selecting a transaction description, searching for an unusual word in it, and pasting a row into a plain text editor. Repeat that test on a second page. A clean match is a reason to try extraction. A selectable logo, a working page number search, or a crisp display at high magnification tells you much less about the table you actually need.
In this guide
Imagine receiving a sixteen-page statement by email. You need the purchases on pages three through fourteen. Before converting sixteen pages and sorting through the results, spend a few minutes testing three lines. Choose one long merchant name, one amount with a decimal point, and one row near a page boundary. Write down what the original shows. You now have known values you can look for in the output, whichever method you use.
What makes a PDF searchable?
An exported report can draw letters at specific positions on the page. A reader can display those letters and often recover them as text. A scanner can make an equally convincing page from a single image. In the image, every visible letter consists of pixels. A text extractor needs some way to relate those pixels to characters. Optical character recognition, usually shortened to OCR, supplies candidate characters from the page image.
There is a third possibility. A scanned page can have a searchable text layer supplied by an earlier OCR pass. Searching and copying may work even though zooming in reveals the edges of a photographed page. You rarely need to reconstruct how the PDF was made. You need to know whether it contains usable text in the relevant areas and whether you can place that text into the correct spreadsheet columns.
A page can contain several kinds of content
A supplier might export a text-based invoice, insert a photograph of a delivery receipt, and add a typed reference in the corner. Searching that reference proves that the reference is searchable. It says nothing about the handwritten quantities in the receipt. Likewise, a bank may provide a searchable account summary followed by images of historical transactions. Test the particular sections you intend to use and note which pages still require attention.
Start with a selection test on an actual data row
Open the document in a PDF viewer and choose its text selection mode if it has one. Drag across a few letters in the middle of a transaction. Some viewers also have drawing, hand, or screenshot tools. A rectangular region that copies an image is a different result from individually selected characters. To distinguish them, paste into a plain text editor. A date, readable words, or an amount can be useful even when the spacing disappears.
Read what you pasted character by character. Suppose a table shows “Harbor Supplies 105.80” but the clipboard contains “Harbor Supplies 10S.8O.” The viewer retrieved text, but two characters are suspect. That file belongs in the category of documents needing review. If the clipboard contains the right words but the amount lands on another line, the characters may be sound while the layout needs reconstruction. Those two findings call for different remedies.
Try a search that tests the body of the table
Find a distinctive word such as “Harbor” rather than searching for “statement.” If the viewer takes you to the exact row, compare the highlighted area with the ink on the page. If the search misses, try a shorter distinctive fragment. A broken word, a line ending, or an unusual space can defeat a full phrase search. Combine that evidence with the clipboard test before concluding that the page lacks searchable text.
Interpret the results without losing the original
If selection, search, and copying all work for the relevant rows, try the bank statement to Excel converter for a workbook or the PDF table to CSV converter for a delimited text file. Keep the source PDF open beside the result. Before editing anything, locate your three known values. Check their order and headings. A correct amount in the balance column instead of the payment column still requires correction.
If nothing in the table selects, look for the original downloadable report. For example, a colleague might have sent a scanned printout of a report that still exists as a digital export. Asking for that export can save both recognition work and hours of verifying uncertain numbers. If the issuer provides a spreadsheet covering the same period and fields, compare the available export with your requirements before spending time rebuilding a printed layout.
If the file is an image-only copy and no better original exists, make a working copy and run an OCR tool you are allowed to use for that document. Select the appropriate document language and review the resulting text on the same lines you tested earlier. Adobe describes adding searchable text in its guide to recognizing scanned PDFs. Keep the original so you can resolve uncertain values from the source image.
Work through a mixed statement
Consider a fictional three-page file for a small design studio. Page one contains the studio address and two summary figures. Page two contains digital purchases. Page three is a scan of an older statement attachment. A quick search for the studio name works on page one, and searching “printer” works on page two. On page three you can read “courier” with your eyes, but selection captures the whole page and copying yields no useful characters.
A first conversion might return many plausible rows. The count alone gives no assurance about the scanned attachment. Record the expected coverage in a short worksheet note. For this example, mark page one as summary, page two as transaction text, and page three as image requiring separate work. If you can obtain a digital replacement for the attachment, use that. Otherwise recognize the affected page in a copy and verify the result before you combine the data.
Decide which sections belong in the final worksheet
The financial summary may show the sum of the digital purchases and the scanned attachment. If you include that total as another transaction, your resulting spend doubles back on itself. Before joining your extracts, label each source page by purpose. Put the summary value in a separate control cell or a separate review sheet. Append only the detailed rows to the transaction table. Leave the source page reference somewhere you can retrieve it later.
Diagnose partial text and unusual symbols
Some documents allow selection but return replacement symbols, empty boxes, or strings of apparently unrelated characters. First try another trusted viewer and compare the same line. If the problem appears only in one viewer, that is useful information for deciding how to proceed. If every viewer returns unusable characters, request a fresh export from the originating application or evaluate an OCR copy against the original. Keep your test narrow until one method produces trustworthy values.
When characters appear correct but the sequence is puzzling, compare a single column with a full line selection. A viewer might recover a complete description and amount separately even though dragging across the line gives a scrambled sequence. That can still be suitable for a positional table extractor. Inspect the actual table preview before labeling the whole PDF defective. A wide gap, a wrapped merchant name, and an intervening currency symbol each create different clues.
Check dates and identifiers separately
A search for “12/05” can locate several rows in the same statement. Add a supplier name or nearby reference to identify the intended one. For identifiers such as 000142, copying into a spreadsheet cell may remove leading zeros through the spreadsheet's own interpretation. Paste into a plain text editor before judging the PDF. If the zeros survive there but disappear during spreadsheet import, you have isolated a downstream formatting problem rather than a missing text layer.
Make the conversion decision reproducible
When several colleagues handle the same type of report, leave a short, specific note with the working copy. For example, record “Pages two and three contain selectable transactions. Page four contains an image of the signed receipt. Only pages two and three contribute transaction rows.” A note like this describes what was checked, where, and for what purpose. A generic statement that the PDF is good provides very little help to the next reviewer.
Before handing off an export, compare the opening data row, a middle row with a long description, and the final row against the original. Then compare the row count or a printed control total if one is available. These checks can reveal a missed page, but agreement on three samples does not establish the correctness of every transaction. Match the depth of review to how the spreadsheet will be used and investigate every consequential discrepancy.
What if a phone lets me copy words from a photo?
Some device features recognize text displayed in photographs or screenshots. Copying that text can succeed while the underlying PDF still contains an image alone. Test inside a PDF reader without invoking the device’s separate image-recognition feature, or examine the extraction preview from the document itself. If a file produces no text there, treat the phone selection as a useful transcription aid and check the actual PDF before describing it to a colleague as searchable.
One last practical test is to repeat the selection after reopening the saved copy in a different viewer. Confirm that you opened the file you intend to convert and that the transaction itself remains searchable. A temporary recognition result shown by one application can be easy to confuse with a text layer saved in the PDF. Give the converter the verified copy and compare the known transaction with the page before processing the rest of the document.