A rumor in my department: the grad students who keep up with the literature are the ones who read during lunch. I do not believe this rumor. What I believe is that they have solved a different problem — they have found a way to move papers through their attention without moving them through their eyes. Chewing a sandwich while staring at Cell Reports is not a sustainable human activity. Listening to Cell Reports while chewing a sandwich, however, is extremely sustainable. That's the whole trick.
This post is about listening to journal articles. Not preprints (those are over here). Not PDFs (those are over here). I mean the actual publisher sites — nature.com, cell.com, science.org, journals.plos.org, pnas.org, frontiersin.org, mdpi.com — where the full text lives behind an Atypon template or a HighWire template and usually a lot of chrome: author affiliation popovers, sharing widgets, eTOC signup forms, "metrics" sidebars, related-article carousels. You want the paper. They want to give you the paper plus everything the marketing team dreamed up at the 2014 offsite.
The extension I use is CastReader. It's free. This post is the tour — one publisher at a time, what the extractor does, and what audio listening actually feels like on each site.
Why Publisher Sites Are Harder Than Preprints
Preprint servers are easy. arXiv gives you LaTeX rendered as HTML. bioRxiv gives you a JATS structure. PMC is essentially a NLM DTD dump of full text. The content is the page.
Publisher sites are the opposite. The content is a minority shareholder in the page. The page is mostly navigation chrome, institutional login banners, Altmetric scores, CrossRef linked citations, newsletter forms, related-paper recommenders, cookie notices, and — on Cell.com specifically — a collapsed <details> panel with author ORCIDs rendered in font-size: 0 text that looks blank but is still technically in the DOM. An older version of CastReader would happily read all of that in a row before ever getting to the abstract. The current version doesn't. Getting to "doesn't" took weeks.
The fix is a rule that sounds obvious in hindsight: if a browser's rendering engine treats something as hidden, a reading tool should too. Collapsed <details>. aria-hidden="true". Zero-size text. Form widgets whose class names contain newsletter | signup | subscribe or whose inputs are of type email | search. All invisible to a normal reader. All should be invisible to the audio. Each fix here was its own small war with some publisher's template choices, and you can feel the payoff when you click a Cell paper now and the first thing you hear is the title and the first author, not an ORCID URL read character by character.
Nature
Dedicated page: /listen-to-nature
Nature is a family. The main journal has the highest bar; Nature Communications and Scientific Reports are the open-access siblings that publish vastly more per year. The template is the same Atypon rendering across all of them, so one extractor handles the whole portfolio — nature.com/articles/s41586-xxxx-xxxxx-x works the same as nature.com/articles/s41467-xxxx-xxxxx-x.
What you hear: title, abstract, main text. Figure and table captions are read as prose. Figure images and table data are skipped because audio can't represent them. Inline citation markers and the References list are stripped so the audio stays readable. Author affiliations and ORCID URLs — gone. Editor's Summary and "At a glance" — kept when they sit near the abstract, because they're useful TL;DRs.
Paywall honesty: CastReader reads whatever your browser has rendered. If your institution gave you access to a Nature subscription paper, the full text is in the DOM and we read it. If you only see the abstract, we only read the abstract. We don't break paywalls; nobody ethically does. For fully open-access Nature content (Nature Communications, Scientific Reports, Nature Portfolio Open Access articles), everything reads end to end.
Cell
Dedicated page: /listen-to-cell
Cell Press runs one of the heaviest Atypon templates I've measured. Author affiliations are nested inside collapsed <details> panels. The "expand authors" button is rendered with zero-pixel font. A newsletter form sits inside the article-wrapper form. Each of those was, at some point, a bug report that sounded like "CastReader reads garbage at the start of every Cell paper."
The current extractor reads title, abstract, main text. Nothing before the title. The Cell Press family — Cell, Neuron, Cell Reports, Current Biology, Cell Host & Microbe, Cell Metabolism, iScience — all share this template and this extractor. iScience and Cell Reports are fully open access; the flagship journals go open 12 months after publication.
If you read Cell Reports regularly, listening is genuinely useful. A typical Cell Reports paper is 4,000–6,000 words of main text. At 1.5x you're through it in 25 minutes, which is one commute.
Science
Dedicated page: /listen-to-science
Science, Science Advances, Science Translational Medicine, Science Immunology, Science Signaling, Science Robotics — all on science.org, all Atypon, all same extractor.
The unique wrinkle on science.org was the Science Adviser newsletter. AAAS liked embedding "SIGN UP FOR THE SCIENCE ADVISER NEWSLETTER" forms in the middle of articles, styled to look like content, and the older extractor would cheerfully read the pitch alongside the Discussion section. The current filter catches any inline <form> whose class name matches newsletter | signup | subscribe or whose inputs capture text (email, search, tel, url). The article-wrapper form that surrounds the abstract — the one containing Save/Cite/Share buttons but no text inputs and no signup class — is preserved, because it's load-bearing structure, not marketing.
Science Advances is fully open access. The main Science and the subscription siblings follow the "full text if you have access" rule. Science News articles (/content/article/...) also read cleanly — they use the same template and sit around 800–1,500 words each, which is perfect for a 1.8x listen.
PLOS
Dedicated page: /listen-to-plos
PLOS is different. It's 100% open access by design — the whole portfolio. PLOS One, PLOS Biology, PLOS Medicine, PLOS Genetics, PLOS Computational Biology, PLOS Pathogens, PLOS Neglected Tropical Diseases, PLOS Climate, PLOS Digital Health, PLOS Water, PLOS Sustainability, PLOS Mental Health. Every paper, every journal, freely accessible.
That means every listen is the full paper. No "abstract only" edge cases, no "please log in through your institution" dead ends. Click play, hear the paper end to end. For me this is the cleanest audio workflow in biomedical reading. PLOS papers also tend to be cleanly written — the journals have house style guides that enforce plain-English structure — so 1.5x is usually fine for the Introduction and Discussion, and 1.2x works for Methods.
If you've ever tried to make sense of "Why Most Published Research Findings Are False" at 2 a.m. under a reading lamp, try it again on a walk with PLOS reading it to you. Different paper.
PNAS
Dedicated page: /listen-to-pnas
PNAS publishes across biology, physics, and social sciences, which makes it one of the widest-scope journals in general science. The PNAS template has one structural feature I love: every research article opens with a short "Significance" paragraph, written for a general scientific audience, that functions as the paper's TL;DR. CastReader reads Significance → Abstract → Main text in that order. You get the gist in the first 60 seconds and can decide whether to continue.
PNAS has a 6-month embargo — articles go fully open 6 months after publication. Inside the embargo, abstracts and Significance paragraphs are public but main text may require institutional access. PNAS Nexus, the newer fully open-access sibling, lives at academic.oup.com and uses a different template; CastReader handles that one through the same listening route.
The most-read list at pnas.org/most-read is excellent raw material for a commute listen. Five papers at 1.5x is about 40 minutes. One week of light reading, compressed into a drive.
Frontiers
Dedicated page: /listen-to-frontiers
Frontiers publishes enormous volume across 50+ journals — Frontiers in Neuroscience, Psychology, Immunology, Plant Science, Public Health, Oncology, and many others. It's 100% open access. The catch is volume: there are Research Topics with sixty-plus related articles, and keeping up by reading is not realistic.
Audio fixes this. Open a Research Topic landing page. Pick the first paper. Listen to it on a walk. When the TTS reaches the end, open the next. One commute gets you through four or five articles from the same topic cluster, which is enough to have an opinion on the trajectory of the work before your PI asks.
The Frontiers template is consistent across the whole family, so the single extractor handles everything. Supplementary materials — separate PDF downloads, datasets — aren't read because they're not in the main DOM, but the main paper reads end to end.
MDPI
Dedicated page: /listen-to-mdpi
MDPI is the volume publisher. Four hundred-plus open-access journals — Sensors, Sustainability, IJMS, Nutrients, Molecules, Cells, Cancers, Energies, Applied Sciences, Materials, Remote Sensing, and on and on. The combined publication rate exceeds any other publisher on this list. You cannot read MDPI. You can only surf it.
Audio is the right mode. Set the speed to 1.8x or 2x, listen to five abstracts from Sensors while making coffee, and pick the one paper that deserves the full listen. Special Issues — MDPI's version of Research Topics — group 20–50 related papers around a theme. Go through the Special Issue titled "Wearable Sensors for Clinical Monitoring" the same way you'd go through an audiobook.
Preprints.org, MDPI's preprint server, uses the same extractor and works the same way.
What About Other Journals
This post covers the seven publisher families with the highest volume and the cleanest payoffs. CastReader's general extractor — the one that handles blog posts and documentation — also works on many other journals: Wiley, Taylor & Francis, Springer, Oxford Academic, Sage, Elsevier (outside of ScienceDirect-specific pages), BMJ, Annals. The Atypon and HighWire templates are extremely common across publishing, so one algorithm covers most of it.
If a journal site reads messily, open the desktop app or the PDF listening page as a fallback — the desktop app reads text selected from any macOS app including Preview, and the PDF tool extracts text directly from uploaded files. But start with the browser extension first. In most cases, the site you're looking at renders HTML that reads cleanly, and clicking the extension button is a one-second intervention between "I should read this" and "I am listening to this."
The Workflow
Here's the practice that stuck for me:
Morning. Open whatever the current publisher backlog is. For me it's an eTOC alert from Nature, a Frontiers Research Topic I subscribed to, and maybe three PLOS tabs a collaborator dropped in Slack. Click CastReader on each. Let them read at 1.5x while I make coffee.
Commute. Put the tabs in listening order. Step out the door. The floating player keeps playing across tab switches and keyboard locks. When I get to the office I have opinions on three papers — which ones were worth the twenty minutes, which ones are going on the "read with eyes" pile, which ones I'm sending back to the requester saying "skim only, not load-bearing."
Afternoon. Desk reading is reserved for the ≤3 papers that earned it. Everything else, I listened to; that counts as having processed it, because I have a position on it.
This is a different relationship with the literature than the "read every paper in your feed" dream. That dream was always a lie. You can process meaningfully more papers if you stop requiring every one of them to earn your full visual attention, and you save the attention for the one or two that are going to shape your work.
The extension is free. No limits, no account, no credit card. Install it, open a journal article you've been putting off, and see how it feels.
Related reading:
- Listen to Research Papers: arXiv, PMC, bioRxiv, OpenReview — the preprint and repository side
- Read PDFs and Papers Aloud — when the paper is PDF-only
- Nature · Cell · Science · PLOS · PNAS · Frontiers · MDPI — individual publisher pages
