ITHAKA / JSTOR Case Study
A 156 Million-Page PDF Accessibility Remediation on AWS
Case Overview
PDF/UA Remediation for a Global Academic Digital Library
PDFix provided the automated structural tagging engine inside ITHAKA’s on-demand PDF remediation pipeline on AWS, replacing Adobe’s Auto-Tag API without a pipeline rebuild.
Industry: Nonprofit / Higher Education & Research (Digital Libraries)
Standards: ADA Title II, PDF/UA (ISO 14289-1), WCAG
Products: PDFix SDK – Automated Structural PDF Tagging
Case Source: AWS Public Sector Blog, Aug 27, 2026
About the Organization
ITHAKA is a not-for-profit organization dedicated to improving access to knowledge and education. Through JSTOR, one of the most widely used digital libraries in the world, it gives researchers, students, and educators access to scholarly content spanning centuries – from single-column journal articles to complex, mixed-media layouts, with source material dating back to 1550.
“Our collaborators at AWS were incredibly helpful in understanding what others were already doing in the accessibility remediation space, thinking through how we could adopt and adapt prior work, and in validating approaches we had thought of toward creating something sustainable in support of our accessibility mission.”
– Dane Hillard, Associate Director of Product Engineering, ITHAKA – quoted in AWS Public Sector Blog
Key Results
98% Accessibility Check Pass Rate
Automated remediation meets PDF/UA validation standards on nearly every document processed.
$0.026 Cost Per Page
Down from the $1-4 per page typical of manual remediation – a 97%+ reduction.
~2-Month Deployment
From a build-and-demo workshop to a live, production pipeline serving real user requests.
The Challenge
Scale, Diversity, and Responsible Stewardship of Nonprofit Resources
Accessibility was already core to ITHAKA’s mission before the revised ADA Title II rule set 2027/2028 compliance deadlines for public-sector institutions.
The math ruled out bulk remediation: at $1-4 per page, converting the collection upfront would cost $156-624 million – far more than most users’ actual needs justify, since each one only wants the single document they’re reading.
The collection itself compounds the problem:
- Three decades of humanities, arts, and social science publishing
- Some born digital, some scanned with patchy OCR
- Everything from plain single-column articles to dense, multi-element page layouts
Key Challenges
- Remediate a massive, growing archive without a multi-million-dollar upfront bill
- Handle wildly inconsistent source quality – born-digital, scanned, and partial-OCR alike
- Meet PDF/UA and ADA Title II standards ahead of a firm regulatory deadline
- Avoid spending remediation effort on content nobody actually requests
- Keep full traceability and audit history for every remediation job
The Solution
On-Demand Remediation, Built on an Open-Source Foundation and PDFix Structural Tagging
ITHAKA built an event-driven pipeline that remediates a document the moment a user requests it. PDFix and veraPDF handle structural tagging and PDF/UA validation, replacing the original Adobe Auto-Tag API without a pipeline rebuild – proof of how modular the architecture is.
How the Automated Workflow Operates
- Event-Driven Initiation – Amazon EventBridge triggers each remediation job with a tracing identifier, so only documents users actually request get processed.
- Parallel Page Processing – AWS Lambda splits each PDF into individual pages, enabling fast, parallel processing regardless of document length.
- Structural Tagging & Alt Text – AWS Step Functions coordinates two tasks per page: PDFix tags the structure – headings, paragraphs, lists, tables – while Amazon Bedrock writes alt text for any images or charts.
- Merge and Validate – Pages are reassembled into a single PDF, validated against PDF/UA standards with veraPDF, and cached in Amazon S3 with a full event log – so the next request for that document is instant.
Solution at a Glance
- PDFix – structural tagging engine (headings, paragraphs, lists, tables) inside an AWS-native pipeline
- Works alongside Amazon Bedrock
- veraPDF – PDF/UA (ISO 14289-1) Validation
- Swapped in for Adobe’s Auto-Tag API with zero pipeline changes
Key Advantages
- On-demand model – no cost spent remediating content nobody reads
- Modular architecture – tools can be swapped or chained as new AI capabilities emerge
- End-to-end tracing for full auditability, from request to output
- Heuristic pre-analysis with human fallback for edge cases automation can’t fully resolve
Results and Measurable Outcomes
Quantitative Impact
| Metric | Result | What it means |
|---|---|---|
| Accessibility Check Pass Rate | 98% | Remediated PDFs meet PDF/UA validation on nearly every run |
| Cost Per Page | $0.026 | Versus $1–4 per page for manual remediation |
| Cost Reduction | 97%+ | Turns a nine-figure problem into an operationally sustainable one |
| Deployment Time | ~2 months | From the February 2026 build-and-demo workshop to production |
| Cost Roadmap | Sub-$0.01 target | ITHAKA is working toward per-page costs below one cent as remediation economics improve |
Operational Improvements
- Remediated PDFs are cached and reused – no reprocessing the same document twice
- Most-requested content gets remediated first, directing effort to where it serves the most users
- A complete event log is stored alongside every remediated PDF for auditability
- Remediation vendors changed (Adobe → PDFix + veraPDF) without touching the rest of the pipeline
- ITHAKA is now applying the same model to new PDFs as they arrive
Why PDFix Works for Academic and Research Institutions
Large academic and research archives share ITHAKA’s exact problem: enormous scale, uneven document quality, and a fixed compliance deadline manual remediation can’t realistically meet.
- Handles extreme document diversity – born-digital, scanned, and partial-OCR PDFs
- Scales through page-level, parallel processing rather than one document at a time
- Drops into an existing pipeline as a tagging engine – no rebuild required to adopt or replace it
- Produces PDF/UA- and ADA Title II-aligned output, validated and auditable via veraPDF
- Brings per-page cost low enough that on-demand remediation is financially viable for a nonprofit budget
- Proven in production alongside AWS-native services
Bring This to Your Own Document Backlog
The same PDFix SDK that powers ITHAKA’s pipeline can automate PDF/UA and WCAG compliance for your own archive – at a fraction of manual remediation cost.
