AutoTag Docling is a local Docker action that uses IBM’s Docling AI engine and the PDFix SDK to automatically tag PDF documents for accessibility, or generate a reusable PDFix layout template JSON. It optionally recognizes formulas as LaTeX/MathML and describes images with a vision-language model – running from a single Docker command with no manual tagging required.
Key Capabilities
Runs locally, fits your workflow
Action Summary
| Free · Local | AutoTag (Docling) | Automatically tags PDF using Docling |
| Free · Local | Create Layout Template (Docling) | Automatically creates layout template json using Docling, saving it as JSON file |
Frequently Asked Questions
What is IBM Docling?
Docling is an open-source AI toolkit developed by IBM Research for document understanding and analysis. It uses the RT-DETR object detection architecture to analyze PDF page layouts, identifying structural elements such as headings, paragraphs, tables, figures, and lists. The PDFix integration uses the Docling model to drive automatic PDF accessibility tagging.
What does auto-tagging mean for PDF accessibility?
PDF tagging is the process of adding a logical structure tree to a PDF document that describes the type and reading order of each content element. Tags allow screen readers, braille displays, and other assistive technologies to correctly interpret document content. Without tags, a PDF is effectively inaccessible. PDF/UA (ISO 14289) requires that all PDFs intended for public use be fully tagged.
Which accessibility standards does this action support?
The AutoTag Docling action produces tag structures that support PDF/UA-1 (ISO 14289-1), WCAG 2.1 Success Criteria 1.3.1 (Info and Relationships) and 1.3.2 (Meaningful Sequence), and Section 508. After tagging, use the VeraPDF Validation action to confirm compliance.
Does this action require an internet connection?
No. After the initial Docker image download, the action runs entirely offline on your local machine. No document data is transmitted to IBM, PDFix, or any third party.
What is the difference between Auto-Tag (Docling) and Create Layout Template (Docling)?
AutoTag (Docling) processes the PDF and writes a complete accessibility tag structure directly to the file in a single step. Use this for one-off documents or when immediate output is the priority.
Create Layout Template (Docling) takes a different approach: instead of tagging the PDF directly, it analyzes the document layout and generates a reusable JSON template file describing the page structure – heading regions, table boundaries, body text areas, footer zones. You can review and adjust this template in PDFix Desktop before applying it. The key advantage is reusability: once a template exists for a given document format, the same template can be applied to every future version of that document automatically, with no repeated work.
This makes Create Layout Template the right choice for recurring documents such as monthly reports, standardized forms, or any document series where the layout is consistent across versions. Learn more about PDFix Layout Templates →
How does this compare to the AutoTag PDF (Amazon Textract) action?
Both actions auto-tag PDFs, but they differ in cost, privacy, and use case. Docling is 100% free, open-source, and runs entirely locally – making it suitable for sensitive or regulated documents with no per-document cost. Amazon Textract is a cloud-based service that requires an AWS account, sends documents to Amazon’s servers for processing, and is billed per page. Textract can offer advantages for scanned or low-quality documents where cloud OCR accuracy is a priority.
