Which errors can PDF improvement fix – and which not (yet)?
The PDF improvement from Eye-Able helps to automatically detect and optimize common accessibility issues in PDF documents. However, there are certain limitations where automatic correction is currently not possible.
What can already be improved
Many common accessibility errors can be automatically detected and improved by the PDF improvement. This already allows numerous issues to be resolved efficiently without the need for manual intervention.
The following accessibility errors can be fixed:
Automatically corrected accessibility errors
| Check step | Explanation | Improvement |
|---|---|---|
| 5 | The document does not contain any PDF/UA identification. | Setting the PDF/UA identification in the metadata. |
| 7.1.3 | The document is not correctly tagged or not tagged at all. | We tag the PDF based on defined rules. |
| 7.1.8 | Missing metadata in the document. | Setting the metadata (e.g. PDF/UA). |
| 7.1.9 | Missing title in the metadata. | Setting the title. Users can set the title. |
| 7.1.11 | The document has no structural information. There is no StrucTreeRoot entry. | Creating a StrucTreeRoot entry and subsequently tagging the document. |
| 7.2.3 | Missing language in the document. | Setting the language. Customers can set the language themselves. |
| 7.18.6 | It is not defined which MIME type the embedded medium has. | Setting the MIME type of the medium. |
| 7.20 | The PDF references external objects (for example graphics). These may not be displayed correctly. | The objects are loaded into the PDF stream and no longer referenced. |
| 7.21.4 | Not all fonts or parts (glyphs) are correctly embedded in the document. | All fonts and parts of fonts are embedded in the document. |
| 7.21.7 | Not all characters of the font have a Unicode encoding. As a result, these characters cannot be read out correctly (for example special characters). | Adding missing Unicode encodings. |
Why can not all errors be fixed automatically?
Some accessibility errors require manual reworking or human understanding of the content.
Reasons for this include:
-
There is not yet suitable logic available for all types of errors.
We are continuously working on being able to fix additional accessibility errors in the future. - Certain errors depend heavily on the content or visual context.
Some errors depend strongly on the content or visual context – here human judgement is necessary. - The PDF improvement generally does not make any changes to the visible structure.
Automatic adjustments could distort content or layout.
Example: An empty heading is not automatically filled with text.
What we are currently working on (planned improvements)
These accessibility errors are planned to be automatically corrected in the future:
Planned automatically corrected accessibility errors
| Check step | Explanation | Planned improvement |
|---|---|---|
| 7.3.3 | Missing alternative text for elements that are tagged as Figure (for example graphics). | Automatic addition of suitable alternative texts. |
| 7.4 | Incorrect order of headings (for example a jump in the heading hierarchy). | Adjusting the heading order. |
| 7.7 | A formula has no alternative text. | Adding the alternative text. |
| 7.18.5 | Links without alternative text. | Adding alternative text. |
Important note
Even if a PDF receives a score of 100% in the PDF checker, this does not necessarily mean full accessibility.
Example:
Scanned PDFs consist purely of image content. They can theoretically pass all automated check steps – but they are not accessible, because their content cannot be read out.
Therefore, manual checks should always be carried out additionally.
Good to know
We are continuously developing the PDF improvement further in order to be able to fix even more errors automatically in the future.