Why Static PDFs Are Undermining Your Research Impact

Systemic Exclusion: Over 75% of scholarly PDFs fail basic accessibility standards, effectively barring blind and low-vision researchers from accessing new knowledge.
AI Invisibility: Static PDFs lack the semantic structure required by modern AI tools, leading to poor parsing and lower discoverability in Large Language Models (LLMs).
Mobile Friction: Fixed-layout documents force users to "pinch-and-zoom" on mobile devices, causing eye strain and reducing engagement.
The Solution: Moving to a "born accessible" HTML-first workflow ensures research is machine-readable and inclusive, while preserving PDFs solely for archival printing. For over three decades, the Portable Document Format (PDF) has served as the undisputed currency of academic knowledge. Created to ensure that a document looked exactly the same on a Windows PC, a Mac, or a laser printer, the PDF solved the "portability" problem of the 1990s perfectly. It allowed the scientific community to transition from physical journals to digital distribution without sacrificing the visual fidelity of the printed page. However, the digital landscape has shifted fundamentally. We have moved from a desktop-centric world to one dominated by mobile consumption, algorithmic discovery, and assistive technologies. In this modern context, the very feature that made the PDF successful—its fixed, immutable layout—has transformed from an asset into a liability. For the modern researcher or publisher, relying solely on static PDFs acts as a bottleneck for research impact. While the format remains useful for the specific "save-as" use case of archiving or printing, using it as the primary vehicle for knowledge dissemination actively limits who can read, understand, and cite your work. The evidence suggests that a strategic pivot to HTML-first publishing is no longer just a technical preference; it is a necessity for future-proofing scientific contribution.

CCM Design TeamMar 22, 2026

The Accessibility Crisis (And Why It’s Getting Worse)

The most immediate and ethically pressing argument against the exclusive use of PDFs is their profound lack of accessibility. While standards like PDF/UA (Universal Accessibility) exist, the reality of academic publishing is one of systemic failure.

The Statistics of Exclusion

A landmark 2024 study analyzing 20,000 scholarly PDFs revealed a staggering gap between compliance standards and reality. The research found that 74.9% of these documents failed to meet even a single accessibility requirement 1. Even more alarming, less than 3.2% of tested PDFs satisfied all key accessibility criteria, such as proper tagging, tab order, and alternative text for images 2.

This data indicates that the "accessibility gap" is not closing; in some sectors, it is widening. Surprisingly, the study observed a decline in PDF accessibility compliance since 2019, particularly among open-access papers 3. This presents a paradox: while the financial barrier to reading research is being lowered through Open Access, the technical barrier for blind and low-vision (BLV) users is rising.

Technical Barriers to Understanding

The failure of the PDF format is structural. Unlike HTML, which is semantic by default (using tags like <h1> for headers and <table> for data), a PDF is often just a visual map of characters on a 2D plane. Without a hidden layer of "tags" added during a costly remediation process, assistive technology cannot interpret the content.

Reading Order Chaos: In multi-column academic papers, an untagged PDF is often read by a screen reader strictly left-to-right across the page. This merges two separate columns into nonsensical sentences, rendering the research unintelligible 4.
The Silent Data: Scientific diagrams are crucial for understanding, yet without proper tagging, they are invisible to screen readers. Many older PDFs function essentially as images of text, creating a complete barrier to access 5.

The High Cost of Remediation

Fixing these issues after publication is resource-intensive. Manual remediation services typically cost between $4 and $10 per page [6]7. For a research institution publishing thousands of pages annually, this creates an unsustainable financial burden. Automated tools often fail to identify complex scientific layouts correctly, requiring human intervention to ensure compliance 8. In contrast, HTML accessibility is largely handled by the content management system—once the template is accessible, every article published inherits that accessibility.

We are currently witnessing a surge in "Chat with PDF" tools, leading some to believe the format is experiencing a renaissance. However, this obscures a deeper technical reality: PDFs are a suboptimal input source for the very AI systems researchers now rely on for discovery and synthesis.

Visual vs. Semantic Data

Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) systems thrive on structured data. PDFs are unstructured. When an AI system attempts to parse a PDF, it must guess the relationship between text blocks. Is this bold text a header, or just emphasis? Is this number a page count, or part of a data table?

Because the PDF format stores characters as coordinate positions rather than semantic meaning, extracting them into a coherent stream of text often results in broken words and lost context 9. This "noise" degrades the performance of the AI model, leading to higher rates of hallucination—where the AI invents facts because it cannot accurately parse the source material 10.

The HtmlRAG Advantage

Recent research into "HtmlRAG" (using HTML for Retrieval-Augmented Generation) demonstrates that preserving the HTML structure of a document significantly outperforms plain text or PDF extraction. Studies show that RAG systems utilizing HTML achieve higher accuracy because the semantic cues (headers, lists, table rows) help the model "reason" about the document's content structure [11]12.

While tools exist to summarize PDFs, they are essentially patches trying to bridge a legacy format with modern machine learning needs. They rely on computationally expensive OCR and heuristic layout analysis that often fail on complex academic documents 13. To ensure your research is accurately ingested, understood, and surfaced by the next generation of AI discovery tools, semantic HTML is the superior format.

The Mobile Friction Point

The "Researcher" persona involves deep reading and analysis, but the context of that reading has changed. Discovery and triage—deciding which papers to read—increasingly happen on mobile devices, where PDFs perform poorly.

The "Pinch-and-Zoom" Fatigue

PDFs are designed to look exactly the same on every device, which means a standard letter-sized academic paper is scaled down to illegibility on a 6-inch smartphone screen. To read a PDF on a phone, the user must zoom in, scroll right to read the line, scroll left to find the start of the next line, and repeat.

This constant manipulation increases the "interaction cost" of the document. Research indicates that reading static, non-reflowable documents on mobile devices leads to higher rates of asthenopia (eye strain) and visual discomfort compared to reading responsive text 14.

Losing the Audience

With over 50% of general web traffic coming from mobile devices, academic publishers who rely solely on PDFs are effectively punishing half of their potential audience 15. When a user encounters friction, they bounce. In the context of research, this means a paper that is difficult to read on a tablet during a commute may simply be skipped in favor of one that is presented in a responsive, mobile-friendly HTML format 16.

The "Black Box" of Analytics

For researchers and institutions, "impact" is measured in engagement. However, the PDF format inherently limits the ability to track and maximize this impact compared to web-native formats.

The Binary Download

Once a PDF is downloaded, the publisher loses all visibility into how it is used. It is a "black box." Did the reader finish the abstract? Did they spend ten minutes studying the methodology, or did they close the file immediately?

HTML analytics provide granular data. Publishers can track scroll depth, time on page, and interaction with specific figures or citations 16. This data is vital for understanding how research is consumed and for optimizing future communication strategies.

SEO and Discoverability

While search engines can index PDFs, they are treated as second-class citizens compared to HTML pages. PDFs often lack the rich, structured metadata (Schema.org) that can be easily embedded in HTML to help search engines understand the content's context, authors, and relationships 17. Furthermore, because search engines prioritize mobile-friendly content, PDFs may rank lower in results than equivalent HTML content, reducing the overall visibility of the research 15.

The Solution: HTML for Distribution, PDF for Print

The evidence does not suggest that PDFs should be eliminated entirely, but rather that their role must be redefined. The current model, where PDF is the primary distribution format, is unsustainable. The future lies in a hybrid model that leverages the strengths of both formats.

The "Born Accessible" Workflow

To maximize impact, research organizations must move away from "creating a PDF and fixing it later" toward "born accessible" workflows. This involves creating content in structured formats like XML, Markdown, or semantic HTML first.

From this single structured source, publishers can automatically generate:

Responsive HTML for screen reading, mobile access, and AI parsing.
Compliant PDF/UA files for users who specifically need to print or archive the document 18.

This approach shifts the investment from expensive post-production remediation to efficient pre-production structuring. It ensures that the primary point of access—the web version—is inclusive, discoverable, and user-friendly, while still providing a downloadable option for those who prefer "digital paper."

By treating the PDF as a secondary export rather than the primary product, researchers can ensure their work is robust enough to survive the transition to an AI-mediated, mobile-first world.

Subscribe to our insights to learn more about modernizing your research dissemination strategy.

Footnotes

1 Semantic Scholar, "Benchmarking PDF Accessibility," 2025. Link Confidence: High 2 ResearchGate, "Uncovering the New Accessibility Crisis in Scholarly PDFs," 2024. Link Confidence: High 3 arXiv, "Uncovering the New Accessibility Crisis in Scholarly PDFs," 2024. Link Confidence: High 4 W3C, "PDF Techniques for WCAG 2.0," n.d. Link Confidence: High 5 Accessibility.com, "PDFs: A Major Barrier to Accessibility," 2023. Link Confidence: Medium 6 DigitalA11Y, "PDF Remediation Services," n.d. Link Confidence: Medium 7 Skynet Technologies, "PDF Accessibility Remediation," n.d. Link Confidence: Medium 8 Allyant, "PDF Accessibility Remediation Software Tools Review," 2021. Link Confidence: Medium 9 Instill AI, "The Best Way to Parse Complex PDFs for RAG," 2024. Link Confidence: Medium 10 ResearchGate, "HtmlRAG: HTML is Better Than Plain Text for Modeling Retrieved Knowledge," 2024. Link Confidence: High 11 MarkTechPost, "HtmlRAG: Enhancing RAG Systems," 2024. Link Confidence: High 12 arXiv, "HtmlRAG," 2024. Link Confidence: High 13 Medium (AI Exploration), "When RAG Meets Document Parsing," 2025. Link Confidence: Low 14 ResearchGate, "Symptoms associated with reading from a smartphone," 2025. Link Confidence: High 15 Shorthand, "The PDF is in terminal decline," n.d. Link Confidence: Medium 16 University of Wisconsin, "Usability and Accessibility Issues with PDFs," 2022. Link Confidence: High 17 CETIS, "Schema.org Briefing," 2014. Link Confidence: High 18 Open Journal Systems, "Digital Publishers Benefit from HTML," n.d. Link Confidence: Medium

publications