Nicole Brennan.

AccessibilityCase study

Assistive technology research.

Finding the barriers screen reader and keyboard users hit on USA.gov, and building the process to fix them.

Back to all work
A research session in progress: a participant’s hands on a braille display keyboard beside a laptop on a table
Role
Research lead; I started it from the ground up
Where
USAGov, GSA
Participants
10 people using screen readers, voice command, braille displays, magnification, contrast settings, and hearing supports
Timeline
2024 to 2025
Partner
The Carroll Center for the Blind

The problem

Technically accessible isn’t the same as usable.

USA.gov reaches an enormous audience, but people using assistive technology were running into complex navigation, inaccessible forms, and unclear workflows. Automated checks weren’t catching it. The only way to find out was to ask, and watch.

How it started

No one in our sessions used assistive technology

I sat in on many USA.gov research sessions. Our screeners asked for people with disabilities, but I never once saw a participant using assistive technology. Our research wasn’t reaching the people who most needed the site to work.

I knew the federal research rules well enough to see a better path. Open-ended generative interviews let us include more participants and recruit through community partners, which meant richer results. So I contacted every assistive technology, blind, and Deaf organization in my region. The Carroll Center for the Blind in Massachusetts said yes. For a small fee, they opened their participant network and helped me write an invitation their community would trust.

How it worked

  1. 01PlanA generative testing plan: open conversation about government websites first, then tasks on navigation menus, forms, and search.
  2. 02Listen togetherIn-person and remote sessions, with teammates from design, engineering, and accessibility sitting in as observers and note-takers.
  3. 03SynthesizeA shared Mural template for notes, theme mapping, and a running barrier tracker, synthesized as a team after every session.
  4. 04Fix and repeatPatterns from the sessions turned into design changes, and the changes went back to participants.

What we heard

Five findings that changed how the team builds.

  • Heading order is navigation
  • Specific beats vague
  • Pop-ups lost focus
  • Sticky nav hurts magnification
  • Moving content is a barrier
  • Dropdowns need better support

Who we tested with

People traditional testing leaves out

Ten people took part, many of them assistive technology users from birth, across a wide range of tools. Every session was mapped on a shared board, so the whole team could see the same barriers come up again and again.

Sticky notes summarizing ten participants by the assistive technology they use: from birth, screen readers, voice command, braille hardware, screen magnifier, contrast, and hearing
Ten participants, by the tools they use.
The Remote AT sessions synthesis board: a column of usability highlights for each participant, color coded, and a summary column of demographics and themes
One column per session, synthesized together.

What we changed

Small fixes, big difference

Most of the changes were small. Together they changed how it feels to use the site without a mouse or without sight.

  • The logo linkClear labeling, one press to the homepage, and skip navigation.
  • The left navigationNo more redundant announcements, and clearer names.
  • Dropdown menusLabeled clearly, the confusing clear control removed, and state abbreviations aligned with Vote.gov.
  • Navigation overallRebuilt for screen readers, with simpler workflows and full keyboard access.

Running the room

Trust is part of the method

Privacy and trust are hard won with this community. They shaped every step, from the invitation to the last question.

  • Time to settle inPeople arrived a few minutes early, so the session didn’t start cold.
  • No surprisesRemote sessions began with just me or a very small team, and I said who would join and why.
  • Consent their wayPeople had different preferences for signing permission forms, so we offered more than one way.
  • Observers out of the wayIn person, one laptop faced the participant so our remote accessibility team could watch without crowding the room.

Two things surprised me: how many participants identified as both blind and Deaf, and how wide the range of braille keyboards was. Both are reminders that “assistive technology user” is not one person.

A user’s action can be a landmark.

What it left behind

This was the federal government’s first assistive technology user research. I built training through GSA Online University so other agencies could run their own, and taught accessibility and neurodiversity workshops to 600+ people across 15+ agencies.

The biggest outcome is a structure for ongoing accessibility testing, so the work doesn’t end when a project does. User feedback now sets accessibility priorities, and the site reached a 91.6% accessibility rating after launch.

Still on the list: replacing confusing toggle language in dropdowns, and clearing up the search results page through a coming Search.gov update.

Accessible names should match the text on screen. When they don’t, screen reader users and sighted users can’t talk to each other about the same page.

Next in IdentityKeyringA decentralized identity wallet that makes the cryptography feel safe.