Skip to content
KovaPDF Research

Government PDF forms that screen readers cannot read

We checked 1,458 PDF forms published by governments in the United States and the United Kingdom, and tested each one the way a screen reader meets it.

Crawled 2026-11-19 to 2026-11-19. 1,458 files. No user files were inspected.

34pass every check we ranof 1,458 files
1,197declare no PDF/UAof 1,458 files
839show no title in the title barof 1,458 files
327have a form field with no nameof 678 files with fields

Results

Six checks, one file at a time

Each check is pass or fail. A file with no form fields is not judged on field names. The table is the whole result in one place.

CheckWhat it meansPassFailNot applicableShare that pass
TaggedThe file carries a structure tree and is marked as tagged. Without it, a screen reader reads the page as one flat block.958500066%
Real text layerAt least 20 characters of real text can be read from the file. A scan with no text layer has almost none.1,43919099%
Language setThe document language is set, so the screen reader pronounces the words in the right language.1,005453069%
Title shownA title is set, and the file asks the reader to show that title in the title bar instead of the file name.619839042%
Every field namedEvery form field carries a tooltip, the short name a screen reader says aloud. One field without one fails the file.35132778052%
PDF/UA declaredThe file declares the PDF/UA standard in its metadata, the promise that it was built to be read by assistive technology.2611,197018%
Passes every checkA file that passes all six, or five when it has no fields.341,42402%

Bases: 1,458 files could be read. Field names are judged only on the 678 files that have form fields.

Key findings

Almost nothing passes everything

Only 34 of 1,458 pass every check

A screen reader user meets all six checks at once, not one at a time. A form that is tagged but has no title still leaves the listener guessing which form is open. A form with a title but unnamed fields still leaves the listener guessing what to type. That is why the bar is every check, and why so few files clear it.

PDF/UA is rare: 1,197 of 1,458 declare none

PDF/UA is the standard that says a file was built to be read by assistive technology. Most government forms never declare it. That does not prove a file is unreadable, but it shows the standard was not part of how the form was made. Only 261 files in the sample declare it at all.

More than half show no title: 839 of 1,458

When a title is missing, or the file does not ask the reader to show it, the title bar shows the file name instead. File names are codes like f1040 or cd-tfa-82. A person using a screen reader, with several forms open, cannot tell them apart by name.

Fields without names: 327 of 678 files with fields

Every field needs a tooltip, the short name the screen reader says aloud. In 327 files with form fields, at least one field has none. The listener hears a blank, or an internal code, where the question should be. The field may still accept typing, but the person cannot know what it asks for.

Charts

Where the failures sit

Pass rates, one bar per check. Text is the easiest check to pass. The standard and the title are the hardest.

Tagged66%958 of 1458 passReal text layer99%1439 of 1458 passLanguage set69%1005 of 1458 passTitle shown42%619 of 1458 passEvery field named52%351 of 678 passPDF/UA declared18%261 of 1458 pass
Pass rate per check, of the files each check applies to. Real text layer 99% (1,439 of 1,458). Tagged 66%. Language 69%. Title 42%. Field names 52% of the files with fields. PDF/UA 18%.

Use this chart

<a href="https://www.kovapdf.com/research/government-pdf-forms-accessibility-2026#chart-checks"><img src="https://www.kovapdf.com/research/government-pdf-forms-accessibility-2026/og.png" alt="Government PDF forms that screen readers cannot read, a study by KovaPDF" width="1200" height="630"></a>
US federal5%22 of 463 pass every checkUS states2%11 of 495 pass every checkUK central government0%1 of 500 pass every check
Share of files that pass every check, by group. US federal 5% (22 of 463). US states 2% (11 of 495). UK central government 0% (1 of 500).

Use this chart

<a href="https://www.kovapdf.com/research/government-pdf-forms-accessibility-2026#chart-groups"><img src="https://www.kovapdf.com/research/government-pdf-forms-accessibility-2026/og.png" alt="Government PDF forms that screen readers cannot read, a study by KovaPDF" width="1200" height="630"></a>

Leaderboards

Who does best, and who does worst

A state or an agency is named only when it has at least 20 files in the sample. Smaller samples are not ranked, because one or two files would move them.

US states, ranked by the share that pass every check

RankStateFilesPass every checkTaggedTitle shownPDF/UAFields named
1New York2110%24%90%10%50%
2California205%100%60%5%100%
3New Mexico205%50%20%5%n/a
4Arizona205%5%5%5%50%
5Utah210%24%24%0%n/a
6Alabama200%15%20%0%9%
7Arkansas200%65%0%0%44%
8Connecticut200%80%5%0%n/a
9Florida200%70%45%30%100%
10Idaho200%5%25%0%89%
11Illinois200%90%30%0%44%
12Indiana200%100%100%0%n/a
13Mississippi200%20%0%0%6%
14Missouri200%50%5%0%21%
15Nevada200%95%45%0%22%
16North Dakota200%90%85%0%50%
17Vermont200%100%35%0%100%
18Virginia200%100%90%0%93%
19Minnesota200%100%70%0%50%
20Nebraska200%90%35%0%9%

Federal agencies, ranked the same way

RankAgencyFilesPass every checkTaggedTitle shownPDF/UAFields named
1Internal Revenue Service9515%97%66%65%21%
2U.S. Department of Labor1173%77%64%56%69%
3U.S. Citizenship and Immigration Services1503%19%6%73%100%
4U.S. Office of Personnel Management990%34%22%0%100%

Federal, state and UK, side by side

GroupFilesPass every checkTaggedLanguage setTitle shownPDF/UA
US federal4635%53%44%37%51%
US states4952%65%72%40%4%
UK central government5000%78%89%51%0%

Why it matters

A form is a task, not a document

Nobody reads a government form for pleasure. A person opens one to file a tax return, claim a benefit, apply for a licence, or register a lasting power of attorney. When the form cannot be read by a screen reader, the person cannot do that task alone.

The usual answer is to ask for help, or to phone, or to visit an office. Each of those costs time, and each takes away privacy. A tax return and a benefit claim hold the details of a life. A form that works with a screen reader lets a blind person keep those details to themself.

The fixes are small, and most of them are made once, when the form is built. A title, a language, a name for each field. The cost of leaving them out falls on the person who can least afford it.

How to fix a form

Seven steps, in the order a publisher should take them

  1. Start from the file the form was made in. Fixes made in the finished PDF rarely last. Fix the source, then publish again.
  2. Give it a real text layer. If the form is a scan, run it through text recognition first, so the words exist as text. Our OCR PDF tool makes a scan searchable.
  3. Tag the reading order. Tag headings, paragraphs and tables in the order a person should read them, so the screen reader does not jump around the page.
  4. Set the language. One line in the document settings, so the screen reader pronounces the words in the right language.
  5. Write a title, and ask the reader to show it. Use the form name a person would recognise, then turn on the setting that shows the title in the title bar instead of the file name.
  6. Name every field. Give each field a short tooltip that says what it asks for, in plain words. Test it with the keyboard alone: tab through the form and listen to what each field says.
  7. Declare PDF/UA, then check with a real screen reader. Automated checks cannot judge reading order or tag quality fully. Listen to the form end to end before it is published.

Method

How we measured this

We crawled public government PDF forms on 2026-11-19 to 2026-11-19, checked each one with the same six automated checks, and counted the results. No user files were inspected. The full list of start pages is in the sources file that ships with this study.

  • Sample: 1,458 files. 463 from five US federal agencies, 495 from 31 US states, and 500 from UK central government. The federal agencies are the Internal Revenue Service, the U.S. Citizenship and Immigration Services, the U.S. Department of Labor, the General Services Administration and the U.S. Office of Personnel Management. The state sample is tax forms first, from each state official tax forms page, with other state forms where the state portal yielded any. The UK sample is form publications found through the gov.uk search.
  • Checks. Tagged: the file carries a structure tree and is marked as tagged. Text layer: at least 20 characters of real text can be read. Language: a document language is set. Title: a title is set and the file asks the reader to show it in the title bar. Field names: every form field carries a tooltip; a file with no fields is not judged on this. PDF/UA: the file declares the PDF/UA standard in its metadata.
  • What was not checked. Reading order, tag quality, colour contrast, whether a tooltip is helpful, and whether the form is complete or correct. Automated checks cannot judge those. A file that passes all six may still read badly, and a file that fails one may still be usable with effort.
  • Limits. This is a sample, not every government form. Some sites return an error to a plainly identified crawler, and some forms pages load their lists with scripts, so those forms are not in the sample. The states are compared on tax forms first, so the state ranking is a tax form ranking, not a ranking of everything a state publishes. The crawl ran on one day, and forms change.

Download the data

Check our work

One row per file: its source, its host, its address, the date it was crawled, each check, its page count and its size. No personal data is in the file.

Download the data (CSV)

How to cite this study: KovaPDF (2026). Government PDF forms that screen readers cannot read. https://www.kovapdf.com/research/government-pdf-forms-accessibility-2026

Most government forms fail the simplest tests. A title, a language, a name for each field. These are small fixes, and they decide whether a blind person can file a tax return or claim a benefit alone.