“AI is reducing options I had used”: exploring faculty perceptions of accessible course design in a ChatGPT world

Synthesis: A mixed-methods study at one Ontario university asked 69 professors about accessibility and Universal Design for Learning at a moment when accommodation demand had outgrown the services designed to meet it: university enrollments in the province rose 17% between 2013 and 2022 while registrations with accessibility services rose 126%. Most respondents knew the vocabulary and believed they already designed accessible courses, but 46% said the institution does not provide sufficient resources, and the qualitative themes include a novel complaint — that Generative AI is undoing assessment options faculty had relied on for Accessibility.

Key Findings

1. Awareness was high, implementation uneven. 77% of 69 respondents were aware of UDL while 21% were unaware, and of 20 commonly used accessibility techniques, four were unfamiliar to more than 20% of respondents.
2. Self-assessment ran ahead of practice. 87% agreed they actively consider accessibility when designing a course and 90% believed they create an accessible learning environment, yet seven of the 20 techniques showed no uptake even among faculty who knew them.
3. Support was the missing piece. Just 46% agreed the university offers sufficient resources on making courses more accessible, with 25% disagreeing and 29% neutral.
4. Assessment variety and pre-posted materials led implementation. 88.41% actively offered varied assessment forms and 88.41% posted slideshows before the lecture, while 13 of 20 techniques reached at least 66.7% active or beginning implementation.
5. The blind spots are concrete and cheap to fix. 30.43% had never heard of OCR-readable PDFs, 26.09% of high-contrast slides, and 20.29% of not relying on color alone to distinguish links or text.
6. Generative AI entered the study as an obstacle. Faculty described AI reversing accessible course design progress — the source of the paper's title quotation — and called for support to keep UDL viable in its face.
7. Recorded lectures were contested. The theme of lecture recordings as an accessibility measure split faculty, and the authors treat the controversy itself as a finding about how faculty wellbeing and accommodation demands interact.

An awareness gap that is not really an awareness gap

The numbers describe a familiar pattern: professionals endorse the goal and mis-estimate their own performance. Nearly 90% of respondents believed they provide an accessible environment while roughly a third had never encountered a technique as basic as producing text-selectable PDFs. The study's framing is generous rather than accusatory — the barrier it identifies is capacity, since fewer than half felt the institution gave them enough resources, and 20 techniques is more than any one instructor can adopt without support. The actionable reading is that targeted lists and short workshops can close specific blind spots that general exhortation to care about accessibility does not.

Why generative AI appears as a setback

The title quotation comes from the study's own faculty interviews, and it captures an interaction the AI-in-education literature rarely reports: the arrival of Generative AI restricted Assessment formats that had been doing accessibility work — take-home tasks, annotated drafting, problem sets whose value came from the process of producing them. Faculty saw the same assignments that lowered barriers for students with disabilities become the ones AI could complete outright, which pushes instructors back toward in-person, invigilated, or more complex assessment. Read this way, AI is not only a technology to make accessible; it is also a force that removes inclusive options from the design space, and the paper argues that UDL support has to be funded and staffed to compensate.

Faculty wellbeing as a design constraint

The fourth theme is that accessibility work is uncompensated and invisible, and that its cumulative load sits alongside a rising accommodation caseload. The authors' recommendations stay inside that constraint: additional resources, forums where faculty can discuss common challenges, better coordination with accessibility services, and recognition for the front-loaded work of designing an accessible course. None of these is a technology fix, which is the point — the study's evidence is about what the institution is willing to spend, not about a tool that could substitute.

What this means for practice

• Instructors. Audit for the four blind spots the data names: text-selectable PDFs, high-contrast slides, color-independent links and text, and physical access to the room. Each is a one-time fix rather than a redesign.
• Faculty developers. Run the survey as an intervention. Participants reported that completing it raised their awareness of techniques they had not considered, which makes a self-audit instrument a cheap workshop activity.
• Administrators. Fund UDL support to compensate for what AI removed. Faculty in this study reported losing assessment formats that had previously worked as accessibility accommodations, and replacing them requires protected time.

Limitations

• One Ontario university, 69 survey respondents, with a self-selected sample of faculty willing to answer questions about accessibility.
• Implementation was measured by self-report rather than by inspecting course sites, so the 87% and 90% agreement figures describe intent and belief, not verified practice.
• The AI-related theme rests on a small qualitative interview set, so it identifies a concern rather than estimates its prevalence.
• The study reports no student data, so the impact of the faculty practices on students with disabilities is inferred rather than observed.

Connected Concepts

• Universal Design for Learning
• Accessibility
• Inclusive Learning
• Assistive Technology
• Neurodiversity
• Generative AI
• Teaching
• Educational Development
• Assessment
• Higher Education

Connected Articles

• LUDIA: A Design and Evidence Statement — LUDIA: A Design and Evidence Statement
• Generative artificial intelligence and the marginalization of minoritized knowledges in higher education — Generative AI, minoritized knowledges and disability
• DysLexLens: A Low-Resource LLM Framework for Analysing Dyslexic Learners Insights from Online Forums — DyslexLens: AI support for dyslexic learners
• Touching and Feeling the Data: A Reusable Software Pipeline for Tactile Statistical Graphs in Accessible Education — Touching and Feeling the Data: A Reusable Software Pipeline for Tactile Statistical Graphs
• From fear to innovation: A case study of transformative faculty development for ethical AI integration in higher education — Faculty development for ethical AI
• The AI Challenge: How college faculty assess the present and future of higher education in the age of AI — The AI challenge: faculty survey findings
• Addressing the Void of AI Policies in Education for Students With Specific Learning Disabilities — AI policies for students with learning disabilities

Citation

Sidhu, G., Atif, F., & Newland, F. (2026). “AI is reducing options I had used”: exploring faculty perceptions of accessible course design in a ChatGPT world. EdArXiv.