Indeed

A continuous research model for accessibility

People with disabilities face higher unemployment rates than the general population — yet many job platforms still fall short of meeting their needs. At Indeed, we had audits and conformance checks, but those only told part of the story. We weren’t seeing how people using screen readers, magnifiers, or voice navigation actually experienced our product, and where things broke down.

I co-led a continuous accessibility research program, partnering with our UX Research Lead. We designed a fast, scalable approach to uncover real usability issues, drive fixes, and shift accessibility earlier in the product development process. The result: more inclusive experiences, better tools, and a stronger accessibility culture across the company.

Assistive technologies

Screen reader user

Screen readers

Person working on a laptop using a magnification tool.

Magnification

A man using a sip and puff device to operate his smartphone.

Alternative input devices

Our approach

We audited the live site experience and identified 11 key tasks across the job seeker journey — from searching and filtering to applying and building a resume. We created simple, repeatable research plans with clear timelines and reporting structures. Each product area received:

  • An Accessibility Usability Score (AUS)

  • Key usability findings

  • Video clips of AT users navigating the flow

  • Actionable recommendations (design, code, or content)

Each round could be completed in just two weeks, enabling us to scale testing across surfaces quickly.

Series of mobile app screens showing job search, profile creation, and application process on a mobile device.

We audited 11 key tasks across employer and job seeker products, mobile and desktop; including Indeed’s most visited pages.

Fixing real problems

We partnered directly with teams to make fixes and improve usability for AT users. We prioritized issues that created confusion or blockers, like missing or redundant alt text, inconsistent labeling, and unclear or absent feedback

We used AUS scores to benchmark progress and track improvement over time and across surfaces.

Screenshot of a digital report discussing usability issues with filters, including quotes from participants and a heading highlighting the lack of filter labels or section headings.
A webpage showing instructions for adding H2 and H3 headings for navigation, with a screenshot of the Indeed job search interface for Support Specialist jobs in Austin, TX, including job details and application options.
A presentation slide with the title 'Filter button name' describing filter buttons in a user interface. The slide contains a blue 'Company Filter Button' with details, a gray 'Within 15 miles' button, and a blue 'Distance Filter Button' with details, all displayed on a light gray background.

Educating and shifting left

To prevent repeat issues, we used video clips and findings to build empathy and drive behavior change. I ran Lunch & Learns, shared examples in our internal channels, and added new guidance to our design system so teams could build more accessibly from the start.

This helped ensure accessibility would be an ongoing practice, not a checklist.

A slide titled 'Fable Insights & Recommendations' with seven numbered tips for accessible design, including adopting a mobile-first mindset, thinking through content, confirming success messages, fixing sticky elements, providing tap options, considering multiple views, and testing with assistive technologies.
A presentation slide titled "Magnification issues ranged from visual defects to outright blockers" with four columns showing different magnification issues: Minor, Major, Critical, and Blocking. Each column contains example images and descriptions illustrating the issues, such as no max-width on images, disabled filter panel zooming, a fixed header obstructing view, and user unaware of changes after selecting a job card.
Screenshot of a user interface with instructions for creating conversations using real words, showing example chat bubbles labeled with names and dialogue, with side notes on features for selecting work history entries and related skills, main text emphasizing starting with story and testing with real words.
Screenshot of a webpage discussing including status messaging in content and design handoffs, with a form for email input showing an error message, and a table explaining error descriptions.

Outcomes

Improved AUS scores across major flows:

  • Search filters: 85

  • Apply: 88

  • Build a resume: 82
    (Industry average baseline: ~65)

  1. Identified 70+ critical issues missed by automation and audits

  2. Created accessibility principles to help guide consistent, inclusive design across Indeed