Images convey essential information in science, technology, engineering, and mathematics communication. Current guidelines on publishing recommend making images accessible to all readers. However, academic publishers do not always follow these guidelines and therefore fail to guarantee access by all readers to the visual content of academic articles. People with severe visual impairments cannot access the visual content of images unless a text alternative describing the images is provided. This study investigates the current use of texts commonly related to images in academic articles, such as captions and mentions, in order to assess their suitability as potential text alternatives to the images for readers who are blind or have severe low vision. A sample of 30 academic articles in the fields of biomedicine, computer science, and mathematics was analyzed and quantitative and qualitative data were collected about images and their related texts. We suggest a practical and sustainable solution that can foster the adoption of good accessibility practices by authors and publishers and facilitate their inclusion in regular publishing workflows.