this looks like something that a trained AI can solve: recognize the meaningful parts of a page and adapt it to the user's needs. just like with every web standard, technology and hype made those accessibility tags obsolete very soon.
This would be a good use case for AI. Unfortunately the implementation right now is pretty poor, at least when it comes to MS PowerPoint. When I add an image to a slide deck, PPT automatically attempts to add a caption with its best guess of what the image is. 90% of the time it says "An image of a cell phone", and I have never once added that image.
