The language of mathematics has always posed a complex problem for students with visual impairments. Accessibility difficulties generally concern: semantic aspects, such as understanding symbols and the statements of theorems and problems; syntactic aspects, such as transcribing mathematical language into accessible languages (e.g., braille); and spatial aspects, such as graphical representation. Over the years, numerous measures have been adopted in terms of accessibility: from the possibility of linearizing formulas with dedicated interfaces, to accessibility standards for major application software and the web (W3C et al.). In recent years, the presence and use of generative AI represents a new frontier: through its use, it is possible to generate alternative texts that help simplify syntactic and semantic decoding and replace the use of images with effective descriptions.

However, there are evident limitations that students and teachers themselves are not yet fully aware of: through this research, we aimed to show the characteristics and limitations of AI in the study of mathematics. With a working group of twenty-five specialized high school teachers, three types of mathematical problems were selected: an algebraic equation containing an error, a Euclidean geometry problem with a construction defect, and a geometry problem that requires different solutions, asking to justify the choices. The questions were proposed to about ten of the most popular and freely accessible AI platforms, and the results were extremely heterogeneous: each platform suggested different interpretations and versions, proposing mathematically unacceptable solutions; some isolated cases highlighted the error, but not in all questions; the same platform provides different interpretations and solutions a few hours after the first submission.

This presents a certainly promising picture for the future, but not yet currently reliable as a study tool for students, blind or not, in high school and university: AI provides partial but not holistic support for the study of mathematics.

Keywords: Technology, special education, mathematics.