Faculty’s digital accessibility assistant: Does generative AI write accurate, meaningful, and useful alt text for course images?

Background Alternative text (alt text) for images is critical for digital accessibility, but faculty often lack time, training, or knowledge to create it. Generative AI (GAI) offers a potential solution, but its effectiveness in educational settings is underexplored. Objective This study evaluated the quality of AI-generated alt text for complex course images from the perspective of faulty teaching online courses. Methods The researchers employed a mixed-methods descriptive design. Ten nursing faculty evaluated 43 alt descriptions generated by ChatGPT-4 for their course images. Participants rated descriptions on 5-point Likert scales for accuracy, meaningfulness for student learning, and usefulness in their course. Qualitative feedback on shortcomings was analyzed thematically. Results Faculty rated the AI-generated text highly on accuracy (M = 4.44, 90.7% positive) and meaningfulness (M = 4.30, 86.0% positive). Usefulness was also high (M = 4.05, 76.7% positive), and all three measures were strongly to moderately correlated. Qualitative analysis revealed three themes of faculty critique: "Accurate but..." (lacking pedagogical context), "Not Quite" (containing subtle semantic errors), and "Too much" (excessively verbose). Conclusion GAI is a useful tool for creating high-quality first drafts of alt text, with potential to save faculty time. However, the findings strongly support the need for a “subject-matter-expert-in-the-loop.” Faculty remains essential for correcting inaccuracies, trimming verbosity, and adding nuance that the pedagogical context requires.