How to Read Research Papers Efficiently
How to read research papers efficiently: the three-pass method with time limits, which sections to read first, and a faster literature review workflow.

To read research papers efficiently, read each one in up to three passes instead of front to back: a 5 to 10 minute scan of the title, abstract, introduction, headings and conclusions; a careful read of the figures and main argument; and a deep, line-by-line pass only for papers central to your work.
Don't start on page one and plow through. A 5 to 10 minute first pass tells you whether a paper deserves any more of your time.
Read the abstract, introduction, headings and conclusions first. Save the methods and the math for papers you will build on, replicate or critique.
In the second pass, study the figures and tables closely and write down terms you don't know. Aim to explain the paper to someone else.
For a literature review, start with three to five recent, highly cited papers. If their related-work sections point to a recent survey, read that first.
Listen to the prose sections, such as the introduction and discussion, and keep equations, tables and code for your eyes.
The fastest way to learn how to read research papers efficiently is to stop reading them front to back. Read each paper in up to three passes, each deeper than the last, and stop as soon as you have what you need. Many papers only ever need the first 5 to 10 minute pass.
Below: the method step by step, the order to read each section in, a literature review workflow, and how to listen to the prose. To hear your PDFs read aloud with natural voices on every device, see Speechly for students.
How to read research papers efficiently: the three-pass method
The most widely shared method comes from computer scientist S. Keshav's short paper How to Read a Paper (ACM SIGCOMM Computer Communication Review, 2007; revised in 2016). Each pass has its own goal, and you decide after each one whether to keep going.
| Pass | Time (Keshav's estimate) | Goal | Stop here when |
|---|---|---|---|
| 1. Scan | 5 to 10 minutes | Decide whether the paper matters to you | It's off-topic or you lack the background |
| 2. Grasp | Up to an hour, for an experienced reader | Understand the argument and its evidence | It's useful but not core to your work |
| 3. Re-create | Many hours for beginners; an hour or two or more for experts | Know it well enough to rebuild and critique it | You can rebuild the paper from memory |
Before you open the PDF, write one sentence on why you're reading it, such as "Does this method work on small samples?" That sentence decides how far you go.
Pass 1: the bird's-eye scan (5 to 10 minutes)
- Read the title, abstract and introduction carefully.
- Read the section and subsection headings, and nothing else in the body.
- Glance at any math to see what theory the paper rests on.
- Read the conclusions.
- Skim the references and tick off the ones you've already read.
After this pass, you should be able to answer Keshav's five Cs:
- Category: What kind of paper is it? A measurement study, an analysis of an existing system, a description of a prototype?
- Context: Which other papers is it related to, and which theories does it use?
- Correctness: Do the assumptions look valid?
- Contributions: What are its main contributions?
- Clarity: Is it well written?
If the answers say "not relevant" or "not ready for this yet," stop. Keshav says one pass is enough for papers outside your area that may prove relevant someday.
Pass 2: grasp the content (up to an hour)
- Read the whole paper with more care, but skip details such as proofs.
- Jot down key points in the margin, plus terms you don't understand and questions you'd ask the authors.
- Study the figures, diagrams and graphs. Are the axes labeled? Do results show error bars? Mistakes like these separate rushed work from excellent work.
- Mark unread references that look important. They're your background reading.
Expect this pass to take longer in a new field. By the end, you should be able to summarize the main argument and its evidence to someone else.
Pass 3: re-create the paper (several hours)
Save this for papers central to your thesis, a replication, a journal club you're leading or a peer review. The goal is to virtually re-implement the work: make the authors' assumptions, rebuild their argument or experiment yourself, and compare. That's how you spot hidden assumptions, missing citations and weak methods.
What order should you read a research paper in?
Read for the claim first and the evidence second. This order suits most empirical papers:
| Section | When to read it | What to look for |
|---|---|---|
| Title and abstract | Pass 1, always | The question, the method in one line, the main result |
| Introduction | Pass 1 | The gap the paper fills and its stated contributions |
| Headings | Pass 1 | The structure of the argument |
| Conclusions or discussion | Pass 1, again in pass 2 | What the authors think they showed, and the limitations they admit |
| Figures and tables | Pass 2 | Whether the data actually supports the claims |
| Results | Pass 2 | Effect sizes, comparisons, what was not significant |
| Methods | Skim in pass 2, in full in pass 3 | Sample, design, controls, anything you'd do differently |
| Related work | For a literature review | Papers to add to your list |
| Proofs, appendices, supplements | Pass 3 only | Details you need to replicate or critique |
Two habits make this faster: look at the key figure before you read the authors' interpretation of it, and keep a running glossary of terms and acronyms for each field.
How to read papers for a literature review
A literature review can mean dozens of papers. Keshav uses the passes to decide what to read, in three steps:
- Find three to five recent, highly cited papers with a few well-chosen keywords in an academic search engine such as Google Scholar. Do one pass on each and read their related-work sections. If they point you to a recent survey or review article, read it and you're largely done.
- Find the shared citations and repeat authors in those reference lists. They're the key papers and researchers. Check where those researchers publish to find the top venues.
- Scan recent proceedings or issues of those venues, then make two passes through everything you've collected. If they all cite a key paper you missed, read it and repeat.
In Google Scholar, the Cited by link under a result shows newer papers that cite it, the fastest way to move forward in time from a classic. Related articles finds similar work.
For arXiv papers, Semantic Scholar's Semantic Reader shows details of cited papers in line as you read. On most English-language arXiv computer science papers, it adds AI-generated skimming highlights labeled Goal, Method and Result. Treat AI highlights and summaries as a map of where to look, not a replacement for the first pass.
How to get through the hard parts
- Unfamiliar jargon: Read a recent review article or textbook chapter before the paper, not after.
- Statistics: Focus on what was measured, how big the effect was and how uncertain it is.
- Math and proofs: In pass 2, read each result's statement and skip the proof. Work through proofs in pass 3, with pencil and paper.
- Unclear writing: It may be the paper, not you. Look for a clearer paper by the same group, such as a later journal version.
How to listen to research papers for free
The introduction, discussion and related-work sections are long stretches of prose that are easy to take in by ear on a walk or commute, which frees desk time for figures and methods.
In Virginia Clinton-Lisell's 2022 meta-analysis of 46 studies, there was no reliable overall difference in comprehension between reading and listening. Reading had a small edge when readers set their own pace, and for questions that require inference. So listen to prose, and read anything you need to stare at. More in listening vs. reading: which is better for comprehension.
Every major platform has a free built-in reader:
- Mac: Choose Apple menu > System Settings > Accessibility > Read & Speak and turn on Speak selection. Select text in any app and press Option-Esc.
- Windows: Open the PDF in Microsoft Edge and press Ctrl+Shift+U to start or stop Read aloud. Use Voice options to change the voice and speed.
- iPhone and iPad: Go to Settings > Accessibility > Read & Speak and turn on Speak Selection and Speak Screen. Select text and tap Speak, or swipe down with two fingers from the top of the screen to hear the whole page.
- Android: Go to Settings > Accessibility > Select to Speak and turn on the Select to Speak shortcut. Swipe up with two fingers (three if TalkBack is on), then tap Play to hear the whole screen.
These work for a few pages, but they read whatever is on the page, including running headers, page numbers and the reference list. For the fix, see how to listen to research papers and skip the references.
How to read research papers with Speechly
Speechly is a text-to-speech app for iPhone, iPad, Android, Mac and Windows that reads PDFs, documents and web pages aloud with natural AI voices. To use it with the three passes:
- Add the paper. Import the PDF, paste a link to the article page, or take a photo of a printed page.
- Check the listening time. Speechly estimates how long each document will take to listen to, which helps you plan a session.
- Pick a voice and speed. Choose from 25+ natural AI voices and set the speed from 0.5× to 2×. Start slower for dense papers.
- Read along for pass 2. Words are highlighted as they're read, and you can jump to any sentence to hear it again. On iPhone and iPad, Speechly can skip repeating PDF headers and footers, so you don't hear the journal name on every page.
- Listen away from your desk. It keeps reading with your phone locked and saves your place, so you can pick up on another device signed in to the same account.
On a Mac, select text in any app, such as the abstract in your PDF viewer, and press ⌘E to hear it in a small floating player. On Windows, press Alt+R. The free plan includes two text-to-speech files with premium voices; Premium is $14.99/month or $139/year for unlimited listening. More on listening to PDFs with Speechly, and on building a research routine in text to speech for graduate students and PhDs.
Troubleshooting
- You read the whole paper and remember nothing. You probably skipped the first pass. Write the five Cs before you go deeper, and close the paper after pass 2 to summarize it from memory.
- Every paper looks relevant. Judge each first pass against the one-sentence reason you wrote down. "Interesting" isn't the same as "useful for this project."
- The PDF won't read aloud or select text. It's probably a scan with no text layer. See how to listen to a scanned PDF.
- You can't concentrate on dense papers. Use short, timed sessions, read along with highlighting, and lower the speed. If focus is a constant struggle, see how text to speech can help with ADHD.
Tips for reading papers every week
- Batch your first passes. Do five or ten scans in one sitting, then schedule second passes only for papers that earned one.
- Keep one note per paper: citation, one-paragraph summary, five Cs and open questions. A free reference manager such as Zotero keeps notes next to the PDF, and those notes become your literature review draft.
- Explain papers out loud. A journal club or a quick summary to a labmate shows where your understanding is thin.
Scan first, go deeper only when a paper earns it, and give your eyes the figures and your ears the prose.
Frequently asked questions
How long should it take to read a research paper?
It depends on how deep you go. In S. Keshav's method, the first pass takes 5 to 10 minutes, the second up to an hour for an experienced reader, and a full third pass can take many hours for a beginner. Most papers you look at never need the third pass.
Should I read the abstract or the conclusion first?
Read the abstract first to see what the paper claims, then jump to the conclusions to see what the authors think they showed. Together they take a few minutes and tell you whether the rest is worth reading.
Is it normal not to understand a research paper the first time?
Yes, especially in a new field. Keshav suggests three options: set the paper aside, come back after reading background material, or push on to a slower third pass. Unfamiliar terms, dense proofs and poor writing slow down experienced researchers too.
How many papers should I read for a literature review?
There's no fixed number; Keshav expects a survey to take tens of papers. Read until new papers keep citing work you already know. Most papers only need a first pass, and a small core gets a second or third.
Can I use AI to summarize research papers?
As a guide to where to look, yes. AI summaries and highlights can miss caveats or get details wrong, so check every claim you plan to cite or build on against the paper itself.
Can I listen to research papers instead of reading them?
For the prose, yes. A 2022 meta-analysis of 46 studies found no reliable overall difference in comprehension between listening and reading. Keep equations, tables, code and figures for your eyes.
How do I take notes on a research paper?
Keep it short: a one-paragraph summary in your own words, the main contribution, the key evidence, its weaknesses and open questions. Write it from memory right after the second pass, then check it against the paper.
Listen with Speechly
Natural AI voices on iPhone, iPad, Android, Mac and Windows, with word highlighting and your place saved on every device. Free to download.
Prices and features of other products were checked on November 18, 2026 and can change. Speechly is not affiliated with the other companies named here.

How to Listen to Research Papers (and Skip the References)
How to listen to research papers and skip the references: 5 ways to cut the bibliography, plus how to stop citations like [12] being read aloud.

Text to Speech for Graduate Students and PhDs
Text to speech for graduate students and PhDs: hear papers and comps lists on the go, proofread your thesis by ear, and start free with Zotero.

How to Read a Textbook Effectively
How to read a textbook effectively in 6 research-backed steps: preview, question, recall, review. Plus what to skip and why highlighting fails you.
