The Best Way to Read Academic Papers

"Read papers faster" is the search; the folder of unread PDFs is the problem. And that folder doesn't grow because you read slowly. It grows because papers are hostile reading objects: two-column layouts, prose interrupted by citations, forty pages where eight matter, and a format designed for printing rather than reading. The real problems are triage, structure, and attention — and it's worth being clear which one a tool actually solves.

Triage: most of a paper doesn't need reading

The honest first step has nothing to do with software: most papers deserve the abstract, the figures, and the discussion, in that order, and only the ones that survive deserve a full read. Where a tool helps is making that surviving set actually get read instead of joining the folder.

Chapter- and section-level summarisation matters more for papers than whole-document summaries, because "summarise the paper" mostly re-derives the abstract — which the authors already wrote. Summarising section 4 specifically, on-device, before deciding whether to read it closely, is triage that saves real time. And because the analysis runs on your own hardware via Apple Intelligence, an unpublished preprint or a paper under review never leaves the device.

Structure: the PDF is fighting you

A PDF is a page-description format, and academic PDFs are the worst case: running headers repeating the journal name on every page, two columns that naive text extraction reads straight across, footnotes erupting mid-paragraph, page numbers everywhere. Velo's import pipeline rebuilds the document on-device before you read a word — sections detected into a navigable table of contents, columns resequenced into actual reading order, headers and page numbers stripped, footnotes extracted and linked to their reference points, OCR artifacts cleaned. Scanned papers go through on-device Vision OCR, with Enhanced Processing (Pro) for the genuinely difficult ones.

The result reads like a book rather than a printout, which is the precondition for everything else: real typography, figures in place, footnotes you can follow and come back from. The full iPad-centric version of this workflow is in reading academic PDFs on an iPad.

Attention: the part no tool admits to

Once a paper is clean, the remaining bottleneck is attention. Papers are exactly the reading that eyes slide off. Velo's answer is Focus Mode (RSVP): one word at a time on a fixed focal point, so attention has nowhere to slide, at a pace you control from 50 to 1,000 words per minute. On difficult material the useful direction is down — at a moderate pace it makes the methods section impossible to skim accidentally, which is the real gain. (What the reading research says about the technique — including its limits — is worth knowing first.)

When you hit an equation, a figure, or a sentence that needs re-reading, one tap drops to Reader mode: the full typeset page, position highlighted, eyes free. Every mode shares one word-level position and switching lands in under 100 ms, so the move costs nothing. Scrub drags you back through the stream when a definition didn't land.

The practical setup

Import from Files, Safari, or Mail via the Share Sheet. Everything is stored on-device and reads offline. Every reading mode, import for 3 documents, stats, and the intelligence dashboard are free; Pro removes the cap and syncs position, highlights and notes across iPhone, iPad, and Mac. Bring the paper you've been avoiding — the forty-page one — and see how much of the problem was never your reading pace.

← All guides