Scanned PDFs: How to Read Them Well on a Phone

A scanned PDF is a book of photographs — page images wearing a PDF extension — and it behaves accordingly: it renders and reads fine, but there's no text underneath, so search, selection, reflow and AI discussion have nothing to grip. Reading scans well on a phone means two things: applying the image-document techniques that make photographed pages comfortable, and knowing honestly which features are structurally unavailable until a text layer exists.

Here's how to read them comfortably, what OCR changes, and the limits stated plainly.

Reading page images comfortably

The comfort toolkit is the PDF-on-phone playbook with extra weight on rendering. Fit-to-width and scroll: scans are usually single book pages, so width-fitting produces genuinely readable text size on modern phones. Zoom that lands back cleanly: old scans have cramped gutters and marginalia — a reader with a fixed zoom ladder (tap back to exactly 100%) beats free-pinching that never quite resets. Dark mode with care: inverting a scan inverts the page photograph — typed text survives it well, but aged-paper tone and illustrations go strange; a good reader inverts hue-aware, and honest advice is to test per-book. Positions: page-plus-offset storage matters even more with scans, since image pages re-render at varying heights; to-the-line restoration is achievable and the mark of a serious reader.

What OCR fixes — and doesn't

Optical character recognition reads the photographs and generates the missing text layer, and where it's available it converts a scan into a searchable, selectable document — sometimes even a reflowable one. The honest caveats: OCR quality tracks scan quality (crisp modern scans OCR near-perfectly; aged, skewed or ornate pages produce error-strewn text), layout complexity confuses reading order (columns, footnotes, marginalia), and the output usually coexists with the images — search works, but what you read remains the photograph. For books that matter to you, the better move than OCR-ing a rough scan is often checking whether a properly digitized version already exists — for anything pre-1930 and notable, it usually does.

The AI limitation, stated plainly

In an AI-companion reader, the text layer is what the companion reads: grounded assistants retrieve from a book's actual text, so a scan with no layer gives them nothing — the book opens, reads and keeps your position normally, but the ask-anything features sit out. This is worth knowing before you wonder if the app is broken: it isn't; the document is silent. Seven Reads handles this case honestly — scanned books are treated as ready and readable, not as errors — and the fix, where you need discussion, is sourcing a text version of the same work.

The triage rule for any scan

Before settling in with a scanned book, thirty seconds of triage: Is this work available as a real text edition (public domain? then almost certainly — take that instead)? If the scan is the only version — personal documents, rare books, annotated copies — is the scan quality worth OCR, or is reading it as pages fine? Scans-as-pages is a perfectly good reading mode for material that only exists that way; the mistake is accepting scan limitations for books that have been sitting in clean EPUB on the free shelves all along.

Frequently asked questions

Why can't I search or select text in my PDF?

It's a scan — photographs of pages with no text layer underneath. Rendering works; anything needing actual text (search, selection, AI discussion) has nothing to grip until OCR generates a layer.

Does OCR make a scanned book as good as a real ebook?

Partially — good scans OCR into searchable text, but the result usually coexists with the page images and inherits scan-quality errors. For notable pre-1930 books, a properly digitized edition almost always exists and beats OCR.

Can AI reading assistants discuss scanned books?

Not without a text layer — grounded companions answer from the book's actual text, and a scan is images. The book reads normally otherwise; for discussion, source a text edition of the same work.