Only 9.5% PAC-compliant: What the new study on PDF accessibility in the public sector reveals

The study significantly expands on previous analyses. It is based on the use of CAAT, a platform for testing digital accessibility, in combination with axesSense, an API-based solution that integrates into existing systems and automatically analyzes large volumes of PDF documents against machine-verifiable PDF/UA and WCAG criteria. Rather than focusing solely on federal ministries, this study examined four groups, referred to as clusters: federal agencies, social security providers, health insurance companies, and municipalities. A total of 76 websites, approximately 929,000 URLs, and 126,880 PDF documents were collected. The actual analysis included 69,944 PDF documents from the period spanning 2018 to the end of 2025. 

This does not provide a snapshot, but rather a reliable overview of how accessible PDF documents are actually made available in the public sector. 

A study of this scale cannot be conducted manually. This is precisely where the importance of automated analysis tools like axesSense becomes clear. axesSense is a REST API from axes4 that evaluates PDF documents in a structured manner based on machine-verifiable requirements from PDF/UA and WCAG; it is comparable to tools such as PAC or axesCheck, but is fully automated and can be integrated into existing systems.

This made it possible not only to determine whether individual documents were compliant, contained errors, or were not tagged at all, but also to identify patterns across tens of thousands of documents. This is precisely what distinguishes individual checks from systematic analysis.

For organizations, this is a key issue: Those that publish a large number of PDFs need not just ad hoc checks, but a scalable overview of quality, risks, and recurring errors across their entire document repository. Only then can PDF accessibility be effectively managed.

At first glance, the trend appears positive. The proportion of PAC-compliant PDF documents rose from 1.5% in 2018 to around 9–12% in subsequent years. At the same time, the proportion of untagged documents initially declined.

However, this progress has not been sustained. For several years now, the proportion of PAC-compliant documents has remained at a low level. Recently, the proportion of untagged PDFs has even begun to rise again. The study therefore reaches a clear conclusion: there has been no fundamental improvement.

This is striking because the study period coincides precisely with a phase in which legal requirements were tightened. Legal pressure alone is therefore clearly not enough to bring about a lasting improvement in the quality of document processes.

Federal agencies at the highest level perform the best. On average, they achieve the highest PDF quality score in the study. At the same time, there is a wide range of performance: some websites deliver good results, while others provide almost no accessible PDFs.

Health insurance companies, social security agencies, and, above all, cities are in a much weaker position. This is particularly relevant because these are digital service points that are especially important to many people in their daily lives, for example, when it comes to health care, pensions, employment, or municipal administrative services.

In other words: In areas where digital accessibility would be particularly important, the quality of many PDF documents remains poor.

The most common errors do not involve obvious layout issues, but rather fundamental technical aspects. These include missing or incorrect document titles, problems with tagged content and artifacts, incomplete metadata, or incorrect language specifications.

For users that can see, such flaws often go unnoticed. For assistive technologies, however, they are critical. This is precisely where a common misconception arises: A PDF can look professional yet still be virtually inaccessible from a technical standpoint.

A look at the author-renderer combinations is particularly revealing. The study clearly shows that good software makes a difference.

For example, the combination of Microsoft Word and axesWord as the renderer achieves over 73% PAC compliance. Other common workflows using Word, InDesign, or general PDF libraries fall well below this figure.

That’s a crucial point. PDF accessibility isn’t a given, but with the right tools, such as axesWord, it’s easy to implement even without specialized knowledge.

The most important insight, therefore, is not merely that many PDFs contain errors. The real insight is this: accessibility is a matter of process.

Many of the errors identified could be avoided with relatively little effort, for example through standardized templates, properly configured exports, or automated validation mechanisms. At the same time, editorial work remains essential, for instance when it comes to alternative text, contrast, or a logical document structure.

The conclusion is clear: If you want to improve PDF accessibility, you can’t wait until the document is finished. What matters most are the templates, systems, processes, and tools used.

The study paints a clear picture: the public sector is still far from meeting a good standard when it comes to PDF accessibility. Only a small fraction of the documents analyzed meet the basic technical requirements. The majority contain errors or are structurally inaccessible.

At the same time, the study also shows that improvements are possible with the right tools. Good results are achieved when PDF accessibility is taken into account during document creation.