The ideals of open access do not exclude the blind and reading-disabled, but the realities of limited funding frequently do. In this work, we survey the 36,000 open-access ebooks and chapters supported by OAPEN to determine the extent of this exclusion and to explore strategies for remediation of poorly accessible ebooks. Only 1144 of these are available in EPUB format, which is more easily used for text-to-speech or braille reading devices than the more common PDF format. Of these 1144 ebooks, we found that only 156 of these contained useful descriptions of images in alt-text attributes. 31% of the images had no descriptions at all, and most of the rest contained single words such as "image" or simply a file name.Rapid advances in machine vision and large language models may provide useful tools towards mitigation of poor ebook accessibility. Our initial experience with these tools suggests that combinations of AI tools together with simple editing and reviewing platforms will provide a cost-effective way forward for the open access community. We are evaluating these tools for automatic description of a set of images and assessing the results. This is a first step to the development of advanced tools to deal with complex tables, graphs and figures which will improve access for all of us, not just the reading disabled.