What this covers
- Selection is the definitive test and takes two seconds
- File size per page is a reliable secondary signal
- Zooming tells you immediately - text stays sharp, a scan softens
- A scan can still be read, printed, cropped, split and compressed
A scanned PDF and a generated PDF look identical on screen. They behave completely differently, and the difference explains most of what people describe as a broken PDF: search finds nothing, text will not copy, extraction returns an empty file, redaction refuses to run.
Establishing which one you have takes two seconds and saves a great deal of time spent debugging a tool that is working correctly.
Test one: try to select a word
Click and drag across a line of text. If characters highlight, the document has a text layer. If the cursor sweeps across and nothing highlights - or a blue rectangle appears over the whole page - it is an image.
This is definitive and it is the only test you usually need.
Test two: zoom in hard
Take the zoom to 400 per cent and look at the edge of a letter. Text drawn from an embedded font stays perfectly sharp at any magnification, because it is being redrawn from an outline each time.
A scan is a grid of pixels. Zoom in and the letter edges become soft, then blocky. There is no ambiguity once you are past about 300 per cent.
Test three: the file size
A generated PDF of text and vector diagrams is remarkably compact - typically 30 to 150 kilobytes per page, and often less. A scanned page is a photograph, and photographs of A4 at a usable resolution run from about 200 kilobytes to several megabytes each.
So a 20-page document of 400 kilobytes is generated; a 20-page document of 30 megabytes is scanned. This is a useful signal when you are looking at a folder listing rather than at a document.
Three tests, three answers
Any one of them settles it.
| Test | Generated PDF | Scanned PDF |
|---|---|---|
| Select a word | Characters highlight | Nothing, or the whole page |
| Zoom to 400% | Letter edges stay sharp | Letter edges go soft, then blocky |
| Size per page | 30-150 KB | 200 KB to several MB |
| Search | Finds matches | Never finds anything |
| Text extraction | Returns the words | Returns nothing |
Checking a document
The reader here reports it directly
Open a document in the reader on this site and the panel underneath tells you how many of its pages carry real text. Extracting the text does the same thing numerically - the result panel states how many pages returned nothing.
The third case: a scan with OCR
A scanned page that has been through optical character recognition carries both: the original image, and an invisible text layer positioned underneath it. Such documents are searchable, which is why some scans behave like real documents.
They have a characteristic tell. Search results highlight slightly the wrong region, because the text layer's positions are a machine's estimate of where the words were. And the text is a machine's reading, so it contains errors - most often a confusion between similar characters that a person would never make.
What you can still do with a scan
A good deal, and it is worth being clear about this because 'it is a scan' sounds more limiting than it is.
A scan: what works and what does not
The dividing line is whether an operation needs characters.
Works normally
- Reading, printing, sharing
- Cropping, rotating, resizing
- Splitting, merging, reordering, deleting pages
- Compressing - and it compresses dramatically, because it is images
- Converting to JPG or PNG
Cannot work
- Searching, which needs characters
- Copying text, which needs characters
- Text extraction, for the same reason
- Redaction, which deletes characters
- Comparing two documents by text
Diagnosing a document in under a minute
- Try to select a word - this alone usually settles it
- Zoom to 400 per cent and look at a letter edge
- Check the file size against the page count
- Run text extraction and read the empty-page count
- If it is a scan, stop trying to search it and plan around that
Questions and answers
Can I turn a scanned PDF into a searchable one?
Only with optical character recognition, which this site does not offer. On clean printed text OCR is very good; on handwriting it is not reliable, and a plausible misreading attached to an exam script is worse than no text at all.
Why is my scanned paper so large?
Because each page is a photograph. A colour scan at 300 DPI is several megabytes per page before compression. Compress PDF typically reduces such a document by seventy per cent or more.
Can I redact a scanned document?
No, and a tool that appears to is drawing a box over an image rather than removing anything. Crop the region out instead, or cover it before scanning.
My PDF is searchable but the results are in odd places. Why?
It is a scan with an OCR text layer. The positions are the OCR engine's estimate of where each word sat, and they are approximate. The text itself will contain occasional misreadings too.
