There is a list of five websites that, between them, account for an absurd share of every academic paper published in English: onlinelibrary.wiley.com, link.springer.com, sciencedirect.com, academic.oup.com, tandfonline.com. If you do a Google Scholar search and follow three random clicks, you will land on one of these five about 80% of the time. These are the plumbing of research.
They are also, each of them, a templating decision made by a platform vendor (mostly Atypon, some HighWire) and then shipped across hundreds or thousands of journals. Which means one well-built extractor solves half of everyone's reading list at once. Which is exactly what CastReader does.
This post is the tour — platform by platform, what the audio actually sounds like, what gets skipped, and what the quirks are. It is the sequel to the journal-publishers tour (which covered Nature / Cell / Science / PLOS / PNAS / Frontiers / MDPI). Together the two posts cover roughly "where do I actually read papers" for most researchers.
Why Lump Them Together
Five sites, five brands, but really one engineering problem.
Atypon — the dominant academic publishing platform — renders Wiley, Taylor & Francis, the American Chemical Society, The Lancet's sister journals, and many more. Their markup is visually distinct across publishers but structurally similar. HighWire (Stanford-born, now part of MPS) renders Oxford Academic and a handful of other journals. ScienceDirect runs Elsevier's in-house platform. Springer Nature runs their own platform too. But all five share the same problem-shape:
A published paper lives inside 4–8 layers of institutional chrome — access banners, ORCID popovers, citation widgets, Altmetric donuts, related-papers rails, download sidebars, supplement selectors, "Article Info" tombstones. The science is a minority shareholder in the page.
The first time you install a reading extension on one of these sites, you usually hear the audio start with "Download PDF" or "Copy citation" or the author's ORCID URL read character by character. It feels like being stuck listening to the marketing team before you can get to the abstract. Fixing this is weeks of work per platform, and the fix is subtractive — the extractor has to know what to delete, not just what to keep.
The payoff of getting the extraction right is audio that starts at the title, moves cleanly through abstract → main text, and never lands on ORCID identifiers or "Get full text access" prompts.
Wiley Online Library
Dedicated page: /listen-to-wiley
Wiley Online Library (onlinelibrary.wiley.com) hosts about 1,600 journals — Angewandte Chemie, Advanced Materials, JCO, Hepatology, and an enormous back catalog across chemistry, materials, medicine, education, and the humanities. It also hosts the full Cochrane Library (cochranelibrary.com), with its Plain Language Summaries that read particularly well in audio.
The Wiley template leans on ORCID and affiliation popovers — clickable spans that expand into author-institution tables. Early extractors would happily read the expanded HTML. The current one treats any role="tooltip" or expandable author popover as chrome and skips it. What you hear is "Paper title. First author, second author, and colleagues. Abstract:" and then the abstract.
Subscription honesty: CastReader reads whatever your institution has loaded. Subscribed = full text audio. Non-subscribed = abstract audio. Gold-OA Wiley articles (Advanced Science, etc.) are always fully readable.
SpringerLink
Dedicated page: /listen-to-springer
Springer Nature — link.springer.com — is the widest catalog on this list. It hosts:
- Journal articles (Nature Communications is dual-hosted here and on nature.com)
- Book chapters (Lecture Notes in Computer Science, Springer Texts in Statistics, countless monographs)
- Conference papers (ICML, NeurIPS proceedings before they migrated)
- Reference works (Encyclopedia of…)
- Protocols (lab method write-ups)
All five render the same way. One extractor handles journal articles, book chapters, and conference papers as one problem. This is genuinely useful: a Springer book chapter is often 8–20,000 words, which is exactly the length where visual reading is a chore and audio is a gift.
What you hear: title, abstract (if present), main text in order. What gets skipped: the "Access options" banner, download links, Altmetric widgets, "Change history," and cited-by rails. BMC journals (BMC Medicine, BMC Biology, etc.) are all OA and read end-to-end.
ScienceDirect
Dedicated page: /listen-to-sciencedirect
ScienceDirect (sciencedirect.com) is Elsevier's main platform — 4,000+ journals. The Lancet lives on thelancet.com and Cell Press on cell.com, but almost everything else Elsevier publishes is here. Current Biology, Neuron, Trends in Cell Biology, JACS, Biomaterials, Tetrahedron — the list goes on forever.
The ScienceDirect template's distinctive feature is the "Recommended Articles" rail and the "Article Outline" sidebar. The Recommended rail is a promotional carousel — CastReader skips it. The Outline is navigation; CastReader skips the outline and reads the actual article sections (Introduction, Methods, Results, Discussion) in their full text.
Open access on ScienceDirect is a mix of gold-OA articles (always readable), green-OA (depends on embargo), and subscription-only. As with the other paywalled publishers, the extension reads what your browser rendered — nothing more.
A specific win for biology researchers: Trends reviews (Trends in Cell Biology, Trends in Genetics, Trends in Neurosciences) are 6–10,000-word syntheses of a subfield, published monthly. They're written at a level that's often the best way to learn what's happening in an area you don't follow closely. And at 1.5x they're a nice 50-minute listen.
Oxford Academic
Dedicated page: /listen-to-oxford-academic
Oxford Academic (academic.oup.com) is the HighWire-powered home of Oxford University Press's 500+ journals. Brain, Bioinformatics, Nucleic Acids Research, MBE, QJE, the American Journal of Epidemiology, Review of Financial Studies, Mind, ELH, Review of English Studies — OUP has deep catalogs in medicine, life science, economics, law, and humanities.
Two things about OUP are worth calling out:
Nucleic Acids Research is 100% gold open access. Every paper, free, full text in the DOM. The annual NAR Database Issue is one of the most densely useful document collections in biomedicine, and listening to it while walking is a genuinely good way to learn what's in the databases your lab should be using.
The humanities journals are long. OUP humanities articles routinely run 8–15,000 words. If you've ever been assigned a review-essay in ELH and felt the physical resistance of sitting down to read it, audio fixes that. Long-form humanities writing is built for linear reading, which means it's also built for linear listening.
Tombstone chrome (publisher metadata band), citation widgets, and supplement indicators all get skipped. You hear the paper.
Taylor & Francis Online
Dedicated page: /listen-to-taylor-francis
T&F Online (tandfonline.com) hosts 2,700+ journals across Taylor & Francis, Routledge, CRC Press, Cogent OA, and Psychology Press. It's Atypon-rendered, sharing much of its DNA with Wiley Online Library — but with even more "trial access" banners and "Related Research" carousels layered over every article.
Routledge humanities and social-sciences journals are the reason most humanities researchers visit this site. Third World Quarterly, Journal of Contemporary Asia, Economy and Society, Environmental Politics, Women's Writing — these journals publish long-form arguments (8–15,000 words is normal) where you need to follow a thesis across many sections. Audio, especially at 1.6–1.8x, is a surprisingly good medium for this kind of writing, because the pace matches how the argument unfolds.
Cogent OA is T&F's open-access imprint. Every Cogent article is free to read and CastReader reads it end-to-end.
The Semantic-Hidden Rule
One theme runs through all five of these sites: the extractor decides what to read by treating browser-hidden content as hidden. This sounds obvious. It is not what most reading tools do.
Specifically, CastReader's current extractor skips:
- Collapsed
<details>panels (unless the user expanded them before clicking) - Anything with
aria-hidden="true" - Zero-pixel-font text (some publishers use this for assistive-tech metadata that should not be read by TTS, which is paradoxically more of an accessibility concern)
- Inline
<form>elements with newsletter / signup / email-input patterns - Author-affiliation popovers that open on click
This rule — if the rendering engine treats it as hidden or non-content, a reading tool should too — is what removes ORCID URLs, newsletter pitches, affiliation dumps, and "Expand author information" from the audio. Each rule was its own small fight with some publisher's template. The cumulative effect is audio that starts at the title, stays on the science, and ends at the last paragraph of Discussion.
Paywall Honesty
All five of these publishers have paywalled content. None of CastReader's behavior changes based on the paywall:
CastReader reads whatever your browser has rendered. Institutional subscribers see full text; the extension reads it. Non-subscribers see the abstract; the extension reads that. Gold-OA articles always read in full. We don't bypass paywalls. Nobody ethically does.
This matters because the user experience is predictable. You don't have to know in advance whether a given paper will read end-to-end — you can just click and the audio will go as far as the text goes. If the extension stops abruptly after the abstract, that's the paywall, not the extension.
The Workflow I've Settled Into
After a year of using this on most of my reading, the pattern that stuck:
Mass triage by ear. Papers I'm unsure about (the ones a colleague dropped in Slack, the citations from a new review) I listen to at 1.8x while walking. Most of them I form a position on within the first five minutes — whether they're worth full attention or whether a one-sentence summary in my own head is enough. Audio collapses the cost of triaging a paper from "20 minutes of focused reading" to "5 minutes of half-attention while moving."
Deep reading with my eyes. The 1–2 papers per week that actually deserve full attention — I read those the old way, at my desk, with a pen. But audio has bought me the ability to reserve that focused reading for the papers that earn it, instead of spending it on the median paper my feed spits out.
Long-form humanities in particular. OUP humanities articles and Routledge essays are where audio has genuinely changed how much I read. 15,000-word essays are exhausting to read with eyes; they're conversational in audio. I've gotten through more long humanities writing in the last six months than in the previous two years.
The extension is free. No account. No credit card. Install it, open any paper on any of these five sites, and see how it feels.
Related reading:
- Listen to Journal Articles: Nature, Cell, Science, PLOS, PNAS, Frontiers, MDPI — the journal-publisher side
- Listen to Research Papers: arXiv, PMC, bioRxiv, OpenReview — preprints and repositories
- Read PDFs and Papers Aloud — when the paper is PDF-only
- Wiley · Springer · ScienceDirect · Oxford Academic · Taylor & Francis — individual publisher pages
