Most PDF annotation is highlighting, and highlighting is among the least effective study techniques that has been formally measured. The large review by Dunlosky and colleagues rated it low utility, and the reason is not mysterious: dragging a yellow bar over a sentence requires recognising that it seems important, which is a much weaker mental operation than restating why.
Worse, a heavily highlighted page is functionally the original page. You have marked 40% of it, which is the same as marking none of it, and rereading the marked parts produces fluency without recall.
So the useful question is what to do inside a PDF instead, given that PDFs are unavoidable in research, law, and study.
The rule: never highlight without commenting
If a passage is worth marking, it is worth eight words saying why. Not a summary of the sentence — your reason.
- "contradicts Baker's finding on p.4"
- "this is the actual claim of the paper"
- "small sample, be careful citing this"
- "use this for the intro"
A highlight says this looked important. A comment says what you thought, and it is the only part that will still be useful in a year. It is also the difference between recognition and generation — see the generation effect.
The practical consequence is that you will highlight far less, which is the point.
A colour scheme, kept small
Three colours maximum, with fixed meanings decided before you start. A workable set:
- Yellow — the author's main claims
- Green — evidence, data, methods
- Pink — I disagree, or something is wrong here
The disagreement colour is the most valuable and the one nobody uses. It marks the passages that will actually generate your own writing, and it makes them findable in a document you skimmed months ago.
Any more than three and you spend attention on colour selection rather than on the argument. See color-coding notes.
Get the annotations out
This is the step that determines whether annotating was worth doing.
Annotations trapped in a PDF are almost as lost as no annotations, because you will not reopen a 40-page document to find one comment. Extraction is what turns them into notes.
Zotero extracts annotations from PDFs into a note in your library, with page links back to the source. For anyone doing research this is the single highest-value part of the workflow, and it is a strong reason to keep PDFs in a reference manager rather than a folder. See how to take research notes.
Obsidian, with a Zotero integration, pulls the same annotations into your vault so they sit beside your own notes.
Preview, Adobe Acrobat and most readers will export a summary of comments, which is clumsier and works.
Then do the important part: convert the extracted comments into your own notes. One idea per note, in your words, standing alone — see atomic notes and literature notes. A list of extracted highlights is raw material, not knowledge.
The format trap
Annotations are stored inside the PDF file, and different tools handle them differently. Two practical consequences:
Standard annotations travel; proprietary ones may not. Highlights and sticky notes written to the PDF standard open anywhere. Some apps store extra data in ways other readers ignore or drop. If you annotate in one app and open in another and things look wrong, that is why.
Always keep a clean copy of the original. Annotate a duplicate. This costs nothing and prevents the situation where a corrupted or over-marked file is the only version you have of a source you need to cite.
If you might migrate later, test the round trip early: annotate, export, open elsewhere, and check that the comments survived. Same principle as export your notes.
Tools, briefly
- iPad with an Apple Pencil is the best experience for marking up documents by hand, and handwriting on the source keeps notes anchored to the material. See how to take notes with an Apple Pencil.
- Zotero for anything academic — storage, annotation, extraction and citations in one place.
- Preview on a Mac is perfectly adequate for basic marking and exports comments.
- Bluebeam is the standard where the PDF is a drawing rather than a document — construction and engineering markup.
- DEVONthink if you are managing a large document library and need OCR and search across all of it — see Obsidian vs DEVONthink.
Scanned PDFs need OCR first
A scanned document is an image. You cannot highlight text that is not text, and search will find nothing.
Run OCR before annotating — most tools offer it — or you will be drawing rectangles over pictures of words. See how to scan documents into your notes.
The minimum viable version
Three colours. Never highlight without eight words saying why. Extract the comments the same week. Write three notes in your own words and let the rest go.
The PDF is where you read. It is not where your notes should end up living.
More in how to highlight effectively, marginalia, and how to take notes.