Finding the Line You Remember but Cannot Place
Somewhere around a third of the way in, someone said something about a lighthouse. You want it for an argument you are having now. You know roughly the tone, roughly the chapter, and nothing else useful.
In a paper book this costs twenty minutes of thumbing. In a reader it costs one search, provided the book is one that can be searched — and the exception to that is worth understanding, because it is not random.
What in-book search does
Open the search bar inside an open book, type a phrase, and get a list of hits with enough surrounding words to tell them apart. Tap one and the book jumps there.
The useful part is the context in the results. Searching a common word in a novel returns forty hits, and forty page numbers tell you nothing. Forty short excerpts let you recognise the one you meant on sight.
Search terms that work better
- Search an unusual word, not the important one. For the lighthouse passage, “lighthouse” may appear thirty times. A neighbouring oddity — a name, a place, a strange verb — appears once.
- Search a fragment, not a sentence. You will misremember the exact words. Three or four in a row that you are sure of will find it; a full sentence from memory usually will not.
- Mind the typography. Books often use curly quotation marks and en dashes. A search containing a straight apostrophe may miss a word your keyboard spells differently from the typesetter.
Where it fails: scanned PDFs
A PDF made by scanning paper contains images of pages. There is no text in it to search, so search returns nothing — not “no results because your phrase is absent”, but “no results because there is nothing to look through”.
The same root cause explains three other things people report as separate bugs: text-to-speech will not read it, you cannot select or copy from it, and reflow mode has nothing to lay out. One missing text layer, four symptoms.
The two-second test is the same in every case: try to select a sentence with your finger. If nothing highlights, the page is a picture.
Fixing it needs OCR, which turns the images back into text. That happens on a computer, not in the reader.
A digital PDF searches fine
Not every PDF is a scan. A PDF exported from a word processor or typesetting system has a real text layer and searches normally. The word “PDF” tells you about the container, not about whether there is text inside it — which is the same confusion that makes people think AZW3 files are locked when the extension says nothing about DRM either.
What search is not for
It finds strings. It does not find ideas.
Searching a philosophy book for “freedom” gives you every occurrence of a word, not every passage about the concept, and the passage you actually want may never use the term. For that, the reliable method is still the old one: highlight as you read, so that later you are searching your own marks rather than the entire text.
Highlights can be exported as a text file, which makes them searchable outside the app too.