Hosted HTML, rebuilt PDF, or manual remediation? Choose the right path

Start with the document's real job

Accessibility is not a file-format contest. A public policy read on a phone has different needs from a signed legal record, a fillable form, or a designed annual report. Start by asking where people will read the content, whether they must download it, and whether the source PDF itself is a controlled artifact.

DocAccessible supports three distinct delivery paths. The first two are included in the software workflow. The third is a human service because it requires document-by-document judgment and repair inside the original file.

Choose hosted HTML for the primary web experience

Hosted HTML renders the document as a responsive web page with real headings, lists, tables, images, and landmarks. It is not a screenshot or a PDF embed. It can reflow for a phone, support reliable keyboard and screen-reader navigation, appear in search, and be corrected in place without distributing a new file.

This route is usually strongest for policies, reports, benefits information, research, course material, announcements, and other public content. It is also useful when the source PDF is scanned or structurally poor and rebuilding its tag tree would still leave a frustrating reading experience.

Publish hosted HTML when

• People will read the content on phones or tablets.
• The content should be searchable and easy to share.
• The document will change and needs simple ongoing maintenance.
• Readers need dependable semantic navigation with assistive technology.
• The source layout is not itself the official record.

Use the rebuilt tagged PDF when the layout can adapt

The automated PDF is a new file generated from the extracted and remediated content blocks. It can include headings, lists, tables, figures, reading order, alt text, language, and document metadata, but it does not edit the source PDF in place. Columns, fonts, spacing, forms, and pagination can therefore change.

This is the practical option when users need a downloadable or offline PDF and the exact original appearance is not required. It is PDF/UA-aligned, not automatically certified. Important files should be checked with PAC or veraPDF and reviewed with assistive technology before formal publication.

Use the rebuilt PDF when

• A downloadable or archive-friendly PDF is required.
• The visual design and page breaks are allowed to change.
• The document is mostly linear content rather than a complex form.
• A human reviewer can confirm the generated result.

The planned conversion-quality pilot measures text retention, reading order, headings, lists, links, images, tables, and routing safety by document class. It does not yet publish an accuracy result. Read the PDF-to-HTML conversion quality methodology.

Use manual remediation when the source must remain intact

Some PDFs cannot be rebuilt without losing something important. Fillable forms, signed records, legal exhibits, certificates, engineering documents, branded reports, and files with precise page references may need to remain visually and functionally consistent with the source.

In those cases, a trained remediator works in the original PDF. They repair the tag tree and reading order, identify artifacts, fix tables and form fields, review alternative text, set document properties, and validate the result. DocAccessible accepts agency projects above 500 pages at $5 per accepted source page, regardless of page complexity. Capacity, schedule, validation depth, and exclusions are confirmed in a written project order, separately from the software plans.

Request manual exact-layout remediation when

• Branding, pagination, or print output must not change.
• The PDF is a form, signed record, legal filing, or archival file.
• Complex tables, diagrams, layers, or annotations need judgment.
• A documented human validation step is required.

Discuss a 501+ page project at $5 per source page.

A practical decision sequence

1. Will people primarily read this on the web? Publish hosted HTML, even if you also provide a PDF download.
2. Does the content need a downloadable PDF? Add the rebuilt tagged PDF when reflow and changed pagination are acceptable.
3. Must the original appearance stay unchanged? Use manual remediation of the source PDF rather than an automated rebuild.
4. Is this a high-impact or formal document? Plan for human validation regardless of the output path.

Many teams should publish more than one format

A government policy can use hosted HTML as the public reading experience and retain a PDF for records. A course guide can use HTML for mobile reading and a rebuilt PDF for offline use. A signed filing can pair a readable HTML alternative with a manually remediated PDF that retains the official appearance.

The important part is to label the outputs honestly. Automated generation and scoring help teams move faster, but they do not promise a pixel-identical source PDF or certify conformance without review.

Standards references

• W3C: applying WCAG 2.2 to non-web documents and software
• Section508.gov: creating accessible PDFs
• PDF Association: Matterhorn Protocol for PDF/UA-1