Skip to main content
Start Free TrialBook a meeting
PDF Tools · AI Auto-Remediation

Untagged PDFs scored 0. After remediation they averaged 96.4

Untagged PDFs went from a 0 to a 96.4 average EqualWeb PDF accessibility report score after automated remediation. In a benchmark of 20 real-world documents across 170 pages, the combined post-remediation average was 96.6, and three already-tagged documents reached a perfect score of 100. With EqualWeb's optional color-contrast correction also applied, the combined average reaches 97.8 and 12 of the 20 documents reach a perfect 100.

  • WCAG 2.2 + PDF/UA
  • OCR up to 400 pages
  • Batch up to 1,000 PDFs
  • No AI training on your files
app.equalweb.com/pdf/remediate
Drop PDFs or browseup to 1,000 files
AI Auto-Remediation
Expert Manual
FilePagesStatusScore
annual-report.pdf42Done94
tax-form-w9.pdf3Tagging...-
scanned-policy.pdf118OCR + tagged88
How it works

What automated remediation does to a PDF

Automatically remediate PDF accessibility issues identified against WCAG 2.2, PDF/UA-1 and PDF/UA-2 checks, then use expert review to verify compliance for the specific document. Upload a PDF and the engine processes headings, tables, alternative text, reading order, fonts, text encoding, contrast, language and metadata. It also supports optical character recognition for scanned files and batch processing for up to 1,000 PDFs at a time.

The benchmark measured the time it takes as well: untagged documents averaged 109.6 seconds per document, and already-tagged documents averaged 71.3 seconds.

These scores are automated report results, not legal compliance certificates. Expert review remains necessary to verify compliance for a specific document.

Try the checker for free
Benchmark

20 real-world documents, 170 pages

The benchmark measured automated remediation across two starting conditions: documents with no existing accessibility tags, and documents that already had some tag structure. Every document in the benchmark improved from its starting score.

Documents with no existing tags10 documents and 104 pages with no accessibility structure at all. They started at an average EqualWeb PDF accessibility report score of 0 and reached 96.4 after automated remediation, with post-remediation scores ranging from 82 to 99. Average processing time was 109.6 seconds per document.
Documents with some existing tags10 documents and 66 pages that already carried some accessibility tag structure. They started at an average EqualWeb PDF accessibility report score of 84.9 and reached 96.8, with post-remediation scores ranging from 93 to 100 - three of these documents reached a perfect score of 100. Average processing time was 71.3 seconds per document.
Combined benchmark resultAcross all 20 documents and 170 pages, the combined post-remediation average EqualWeb PDF accessibility report score was 96.6. The 84.9-point difference between the two groups' average starting scores narrowed to 0.4 points after automated remediation. With EqualWeb's optional color-contrast correction also applied, the combined average reaches 97.8 (12 of 20 documents reach 100); see the FAQ below for the full picture, including the one document it did not improve.

Read the full benchmark case study - methodology, page counts, starting conditions and before-and-after results for all 20 documents ->

Note: Scores reflect the EqualWeb PDF accessibility report score (0-100), based on 80+ automated checks across 8 categories. A higher score reflects fewer detected accessibility issues; human review is recommended to confirm compliance.

Remediated automatically

Structure assistive tech can read

Accessible PDFs are about structure, not looks. Where no tag tree exists the AI builds one, then repairs headings, tables, lists and reading order - and keeps going below the surface: font embedding and Unicode repair so screen readers speak the real text, WCAG AA contrast, document language and metadata.

Complex tables or legal docs? See Expert Manual Remediation ->

No AI training on your files, ever - see Security & Data Privacy ->

H1
alt
Table
Order 1-9
Auto-remediation at scale

Hundreds of PDFs, one upload

From a single form to a back-catalogue of reports - tag them all to standard, automatically.

Headings & reading orderCorrect heading levels and a logical, tagged reading order.
Tables & form fieldsHeader associations and labelled fields, tagged correctly.
AI alt textDescriptive alternative text generated for every image.
OCR for scanned filesTurn image-only PDFs into tagged text - up to 400 pages.
Batch up to 1,000 PDFsUpload a whole library and remediate it in one job.
Checker report includedA before/after score and report ship with every remediation.
Fonts & text encodingFont embedding and Unicode repair, so screen readers speak the real text and it stays searchable and copyable.
Contrast, language & metadataWCAG AA color contrast, document language and title, plus navigation bookmarks - set automatically.
AI output you approveAI-written alt text and labels are recommendations - reviewed and approved next to a screenshot of the exact region.
What a score does and does not mean

Does an automated score confirm compliance?

No. An EqualWeb PDF accessibility report score is the output of an automated checker that runs 80+ checks across 8 categories. It is not a legal compliance certificate.

A higher score means the automated report detected fewer accessibility issues. The automated engine addresses detected structural issues, including reading order, table headers and document language. This remediation helps you achieve compliance by resolving structural barriers detected by the automated checker.

Expert human review is still required before making a compliance determination for a specific document. Complex data tables, nuanced reading orders and context-specific alternative text require human judgment to evaluate the experience for people using assistive technology.

In this benchmark, automated remediation produced a combined average EqualWeb PDF accessibility report score of 96.6. Expert review remained necessary to verify compliance with WCAG 2.2, PDF/UA-1 and PDF/UA-2 for each specific document.

Complex tables or legal documents? See Expert Manual Remediation ->

FAQ

AI PDF auto-remediation - frequently asked questions

How does AI auto-remediation for PDFs work?
You upload a PDF to EqualWeb's AI Auto-Remediation and the engine remediates the document in one pass: it builds the tag tree where none exists, sets heading levels, table headers, image alt text, lists and a logical reading order, repairs font embedding and text encoding so screen readers speak the real text, corrects color contrast, and sets the document language, title and metadata - aligned to WCAG 2.2 and PDF/UA. The output is a tagged, accessible PDF you download, together with a Checker report and an accessibility score showing what was fixed.
What does the AI actually fix in a PDF?
EqualWeb's AI Auto-Remediation adds the semantic structure assistive technology needs - heading levels, a corrected logical reading order, table markup with row and column headers, properly structured lists, descriptive link tagging, AI-generated alternative text for images and tagged form fields - and then goes below the surface: font embedding and Unicode repair so text is read aloud correctly and stays searchable, WCAG AA color contrast, navigation bookmarks, and document metadata such as title and language. Accessible PDFs are about structure rather than looks, and this is the layer screen readers read.
How many PDFs can I remediate at once?
EqualWeb's AI Auto-Remediation accepts batch uploads of up to 1,000 PDFs in a single job. You can drag and drop files, browse, or upload from a URL, and the system flags scanned, image-only documents automatically, running an extra OCR step that handles files up to 400 pages. Every file in the batch is tagged and comes back with its own accessibility score and Checker report.
What languages does AI PDF remediation support?
EqualWeb's automatic AI tagging is currently optimized for English and Latin-script languages. PDFs in other languages can be uploaded, but automated tagging support may be limited. For documents the AI cannot fully handle, EqualWeb's IAAP/CPWA-certified experts can remediate the file by hand through Expert Manual Remediation and verify it with real screen readers.
Can I upgrade an AI-remediated PDF to expert remediation?
Yes. Every file processed by EqualWeb's AI Auto-Remediation can be sent on to certified experts using the Upgrade to Manual option in the dashboard. IAAP and CPWA-certified accessibility specialists then review and re-tag the document by hand, delivering a higher-quality, human-verified remediation. Use it when an AI-remediated file needs deeper quality work or formal verification.
Is AI trained on my documents?
No - never. EqualWeb does not train or fine-tune AI models on client documents, and does not retain documents, text fragments or images for training. Most structural fixes are performed by EqualWeb's own deterministic remediation engine; AI is applied only at defined points, such as image descriptions and form labels, under minimal-exposure rules. File retention is under your control - 0, 90, 360 days or never - and the Security and Data Privacy page details the full commitments.
What kind of accessibility score improvement can I expect?
Results vary by document. In this benchmark, every document's EqualWeb PDF accessibility report score improved. Files that started with no tags at all improved to an average EqualWeb PDF accessibility report score of 96.4, with post-remediation scores ranging from 82 to 99. Documents that already had some tags improved to an average of 96.8, ranging from 93 to 100 - three of those documents reached a perfect score of 100. The combined average across all 20 documents and 170 pages was 96.6. EqualWeb also offers an optional color-contrast correction, switched off by default because it can change how branded text looks - for example, recoloring a heading from white to near-black. Applied to the same 20 documents, it raised the combined average to 97.8 and the perfect-100 count to 12 of 20, though not for every document: one document scored lower with the correction on, and two were unchanged. Every remediation includes a Checker report running 80+ automated checks across 8 categories so you can review the issues the checker detected. These benchmark results do not guarantee the same score for every PDF.
Does an automated score guarantee legal compliance?
No. An EqualWeb PDF accessibility report score measures what the automated checker detects. Automated remediation helps you achieve compliance by addressing detected structural issues, but it does not certify that a document complies with every applicable accessibility standard or legal requirement. Expert manual review is still required before making a compliance determination for a specific document.
How long does automated remediation take?
In the 20-document benchmark, average processing time was 109.6 seconds for untagged documents and 71.3 seconds for already-tagged documents. The 109.6-second result is about two minutes per document. These figures describe averages measured in that specific sample, not guaranteed processing times. Results can vary based on document length, complexity, and whether optical character recognition is required.
Remediate PDFs

Start with one PDF

Score one of your PDFs with the free checker. Review its current accessibility score and detailed report before deciding whether to apply automated remediation.

Try the checker for free

Last updated

Talk to a specialistBook a meeting