The PDFs linked from a public-facing website now carry two obligations at once. Inclusion under accessibility law for the human visitor, and AI readiness for the systems finding, interpreting, and acting on the same content externally. The market that answers that dual demand splits into four broad approaches, each with a different output philosophy and a different answer to each half of the obligation.
Category 1. PDF-to-HTML conversion
Vendors in this category convert PDFs into inclusive web pages or HTML transcripts. The source PDF is not modified. The output is a browser-native experience served alongside, or in place of, the original document. This category aligns with the HTML-first direction from Section508.gov and GOV.UK Government Digital Service.
Vendors include aicm, DocAccess, TestParty, Pneuma Solutions Scribe, and DocAccessible.
Pros. Direct alignment with HTML-first regulatory guidance. Structural legibility for AI retrieval systems, because HTML is what those systems read natively. The source PDF is preserved, which keeps point-in-time records, signed documents, and third-party-authored content intact. The approach scales to the full estate without per-document manual effort.
Cons. Where a buyer or procurement team reads the legal requirement as requiring the PDF itself to meet PDF/UA, this category does not produce that artefact. A live legal debate sits around whether an accessible HTML alternative satisfies the obligation in full, or operates under equivalent facilitation provisions. Buyers should take their own legal advice on which interpretation applies to their circumstances.
Category 2. PDF remediation to PDF/UA
Vendors in this category modify the PDF itself. Reading order, tag structure, alt text, metadata, and form field labelling are added or corrected so the source file meets PDF/UA and WCAG criteria. Automation is used to accelerate the work, with a human review layer for complex content in most vendors.
Vendors include Recite Me, ConvertAPI, tabnav, Eye-Able, EqualWeb, PDFix, Equidox, and GrackleDocs.
Pros. Produces a PDF/UA artefact where procurement or legal interpretation requires one. Preserves the PDF-as-document user experience for readers who prefer to download the file. Mature vendor landscape with multiple established products.
Cons. The output remains a PDF, which is structurally opaque to AI retrieval systems regardless of how well it has been tagged. Per-document processing scales linearly, so a large estate remains a cost-and-time function of estate size. Where the source document cannot be altered for legal or evidential reasons, including signed records, third-party documents, statutory notices, and point-in-time records, modifying the PDF is either unavailable or creates questions about document integrity.
Category 3. Hybrid enterprise and managed services
Vendors in this category combine automation with human quality assurance through a service portal, with per-page pricing and managed delivery. They have strong presence in regulated verticals including financial services, healthcare, and transactional communications.
Vendors include Crawford Technologies, Allyant, Continual Engine PREP, Apex CoVantage DocumentA11Y, AbleDocs which is now part of the GrackleDocs group, and Level Access LevelDocs.
Pros. High-confidence output for complex content where automation alone is insufficient. Strong security posture across the category, with SOC 2 Type II, HIPAA, PCI-DSS, and ISO 27001 certifications appearing commonly. Suitable for regulated vertical programmes with service-level agreements and audit artefact requirements.
Cons. Commercially closer to a professional service than a product. Procurement and implementation timelines typically run in weeks to months rather than days. Per-document cost is higher than pure-automation alternatives, and the output inherits the category 2 limitation on AI readiness where the deliverable is a tagged PDF rather than an inclusive web page.
Category 4. Developer APIs, SDKs, and QA tooling
Vendors in this category sell to engineering teams building in-house accessibility pipelines, and to the vendors in the other categories. The output is a technical component rather than a finished deliverable for a business owner.
Vendors include Adobe Acrobat Services PDF Accessibility Auto-Tag API, PDFix SDK, and callas software pdfGoHTML.
Pros. Reduces engineering lift materially for organisations building their own pipelines. Cost per document at scale can be low. Developer-owned workflows give the implementing team control over integration points and output format.
Cons. Requires in-house engineering capacity that most public sector and traditional corporate functions do not carry. Neither discovery nor delivery is solved by the component itself. Quality assurance sits with the implementing team, and the approach is not suited to a business owner who needs the problem retired rather than a new pipeline to run.
An idea of what this costs
Manual remediation at 23 minutes per page, at $50 per hour in the US and £40 per hour in the UK, gives a baseline worth working through on a realistic estate. For 1,000 PDFs averaging 9 pages each, aggregate work is approximately 9,000 pages or 3,450 specialist hours. At the US rate, that is approximately $172,500 in direct cost. At the UK rate, approximately £138,000.
These figures cover execution only. Discovery, triage, prioritisation, and version management sit outside them and add further cost that varies with the state of the estate.
AAAnow / Sitemorse internal data indicates automation delivers 85% or more improvement in accessibility compliance compared with the established manual model, and is delivered in hours rather than months. This is an internal figure, labelled as such, and it describes improvement against the manual baseline rather than claiming any specific compliance outcome for a specific estate.
The commercial point that sits alongside the numbers is direct. Manual cost is a cost against a single processing of the estate. It gets paid again each time the document is revised, a new document is added, a supplier publishes something new, or an old version needs replacing. Automation-first cost is a cost against the estate as a living system.
The archive question
Most regulatory frameworks provide for an archive category, where content that is no longer in active use and is clearly marked as such may be treated differently from live content. In the US, 28 C.F.R. 35.200 provides for undue-burden relief, 35.203 provides for equivalent facilitation, and 35.205 addresses minimal-impact non-compliance. In the UK, the disproportionate burden assessment under PSBAR 2018 performs a similar function. Archive status may, in the right circumstances, reduce or remove the inclusion obligation for content inside it.
The AI readiness obligation is not reduced by the same mechanism. Archived content that remains visible to crawlers, indexable by retrieval systems, and reachable by language models continues to shape the external picture of the organisation. Archive as a regulatory category does not put content outside the AI readiness perimeter. It moves the content into a section of the estate the organisation is less actively managing.
The decision point is the trade-off. Content moved to archive to reduce inclusion cost may continue to represent the organisation externally without active governance. Where the archived content is factual, currently accurate in its own terms, and contains no misleading or superseded material, the trade-off may be acceptable. Where it is not, the archive decision buys relief on one obligation at the cost of a weaker position on the other.
Reading this alongside the rest
The four categories, the cost contrast, and the archive trade-off all assume the content stays where the organisation put it. What happens when a PDF leaves your site picks up at the point when it does not. The two halves of the obligation the categories are tested against are set out in full in The two obligations on your PDFs. The operational stages the market responds to, discovery through version management, are covered in The five stages of dealing with a PDF estate.