Study · October 2026
How accessible are public PDFs in Austria?
We checked 193,198 publicly available PDFs from 382 public institutions against the machine-checkable rules of PDF/UA-1 (ISO 14289-1), the international standard for accessible PDF. 5.8% pass: about 1 in 17.
- PDFs checked
- 193,198
- public institutions
- 382
- pages
- 4,262,431
- pass PDF/UA-1
- 5.8%
Where every PDF lands
Each square is 1 % of all PDFs checked. Just under half are tagged but break at least one rule, and almost as many have no structure at all. 4,441 tagged files fail only on metadata such as the title, which can be fixed without touching the content. 10,133 files have no text on their first pages, mostly scans.
- 5.8%pass the PDF/UA-1 check
- 48.6%are tagged, but break at least one rule
- 45.6%have no tags at all
Newer files do better, but not by much
By creation year. Before 2016 almost no PDF passes. Since 2010 the share of tagged files has roughly doubled; the share that passes peaked at 9.5% in 2024 and stands at 7.3% for 2026, which runs to early October.
| Year | PDFs | Tagged | Passes PDF/UA-1 |
|---|---|---|---|
| 2010 | 2,216 | 30.4% | 0% |
| 2011 | 2,808 | 39.7% | 0% |
| 2012 | 3,117 | 34.2% | 0% |
| 2013 | 3,749 | 37.2% | 0% |
| 2014 | 4,757 | 40.9% | 0% |
| 2015 | 5,210 | 46.5% | 0.1% |
| 2016 | 6,257 | 52.6% | 3.1% |
| 2017 | 6,210 | 52.7% | 2.1% |
| 2018 | 9,163 | 58.6% | 2.2% |
| 2019 | 9,678 | 56.4% | 3.8% |
| 2020 | 11,319 | 55.2% | 5.6% |
| 2021 | 12,860 | 62.2% | 7.6% |
| 2022 | 14,095 | 60.1% | 8.1% |
| 2023 | 15,817 | 58.6% | 8.3% |
| 2024 | 18,660 | 63.3% | 9.5% |
| 2025 | 23,904 | 61.6% | 8.7% |
| 2026 | 31,933 | 58.8% | 7.3% |
Tagged is not the same as passing
By authoring tool, from the PDF metadata. Office exports are almost always tagged, but tags alone rarely get a file through. Printed and scanned PDFs are almost never tagged.
By kind of institution
Pooled across each group. 148 of 324 institutions with at least 20 PDFs checked have not a single file that passes.
The rules broken most often
Share of all PDFs checked that break each rule; most files break several. Metadata tops the list. Behind it, many files lack the structure itself, and tagged files fail too, on fonts, links and images without alt text.
- Title not shown70.2% of PDFs
The viewer shows the file name instead of the title, so a screen reader announces something like “scan_0042.pdf”. PDF/UA-1 7.1-10
- No PDF/UA identifier67% of PDFs
The file doesn't declare that it targets PDF/UA. Invisible to readers, required by every audit. PDF/UA-1 5-1
- Content neither tagged nor artifact53.7% of PDFs
Parts of the page are neither structure nor marked as an artifact. Screen readers skip them, or read headers and footers mid-text. PDF/UA-1 7.1-3
- Not marked as tagged44.5% of PDFs
The PDF doesn't report itself as tagged, so assistive technology treats it as unstructured. PDF/UA-1 6.2-1
- No structure tree, no tags42.2% of PDFs
No headings, lists or tables. Screen readers read one undivided stream of text with no way to jump. PDF/UA-1 7.1-11
- Fonts not embedded38.6% of PDFs
The fonts are missing from the file. Characters may look different or be read out wrongly. PDF/UA-1 7.21.4.1-1
- Text language not set35.8% of PDFs
Speech output doesn't know the language, so German text may be read with English pronunciation. PDF/UA-1 7.2-34
- No title in the metadata30.6% of PDFs
Without a title, tabs, lists and search results show only the file name. PDF/UA-1 7.1-9
- Annotations without description29% of PDFs
Comments, stamps and similar annotations have no description, so assistive technology can't say what they are. PDF/UA-1 7.18.1-2
- Links without description28.9% of PDFs
Screen readers announce just “link” or the raw address, not where it goes. PDF/UA-1 7.18.5-2
- No XMP metadata23.9% of PDFs
Title and PDF/UA identifier have nowhere to live in the file. PDF/UA-1 7.1-8
- Figures without alt text23.1% of PDFs
Blind readers never learn what an image, chart or logo shows. PDF/UA-1 7.3-1
What a checker can't see
A pass here means no rule a machine can test was broken. Whether alt text says what matters, or headings form a real hierarchy, is beyond veraPDF. doclarity adds an AI review for exactly those questions, and a person reviews the result and has the final word.
Check your own PDF for freeMethod and limits
- Who
- 382 public institutions in Austria in twelve categories: ministries, supreme bodies, courts, federal agencies, federal states, cities, social insurance and health, universities and research, chambers, state-owned companies and culture.
- How the files were obtained
- Publicly available PDFs, retrieved automatically from each institution's own websites, starting at the home page and the sitemaps. robots.txt and crawl-delay were honoured, with at most 2.5 requests per second per host. Sites that turned the retrieval away were not worked around (24 institutions).
- Limits
- At most 2,000 PDFs and 40 minutes per institution. 156 institutions reached a limit, so their figures are a sample; 202 were covered completely. Files over 60 MB were not checked.
- The check
- veraPDF, profile PDF/UA-1 (ISO 14289-1), which tests only the machine-checkable rules. “Pass” is an upper bound: reading order and the quality of alt text are left open.
- The files
- Each PDF was deleted straight after checking. Only check results and file metadata were kept, and this page shows totals only, never single institutions.
- When
- October 2026. Files created in 2026 run to early October.
Cite as: Klartext AI, “How accessible are public PDFs in Austria?”, doclarity study, October 2026, https://doclarity.eu/study