AI hallucination

The generation of plausible-sounding but factually incorrect or fabricated information by AI systems, particularly large language and multimodal models. In accessibility applications, AI hallucinations are especially dangerous because users who cannot independently verify visual information must trust the AI's descriptions. Examples include describing objects that are not present, providing incorrect spatial directions, reading labels that do not exist, or defaulting to general knowledge about what a product "typically" contains rather than actually reading the specific label shown. For blind and visually impaired users relying on AI for real-world navigation and object identification, hallucinations can lead to confusion, wasted effort, and safety hazards.