Feed a model a website full of missing alt text and unlabeled buttons, and guess what it spits back out. More of the same. That’s the argument AudioEye lays out over at MarTech: AI now writes and builds a big chunk of the web, and it learned from a web that was never built for screen readers in the first place. So the same gaps get copied, pasted, and shipped at scale.
The screen-reader user hits a wall either way. The twist is who’s holding the fix. Not the dev team buried three sprints deep. Marketing. The people picking the tools, approving the pages, and signing off on what actually goes live. AudioEye’s point is that if AI is generating your customer experience, someone on the marketing side has to check whether that experience works for the folks who can’t see it. Otherwise you’ve just automated the problem.
About 1 in 4 U.S. adults has a disability, and a chunk of them can’t use a page that AI slapped together without alt text, keyboard navigation, or readable contrast. If you run a site or a store, that’s paying customers who bounce, plus the roughly 4,000-plus ADA web lawsuits filed a year that name businesses of every size. The opening: the tools writing your pages can also fix the gaps, if someone tells them to, and most competitors won’t bother.
The pitch was that AI would build the web faster and cheaper. Nobody mentioned it would copy every broken button and unlabeled image along the way, at scale, in seconds.
Turns out a model trained on a web that failed 96% of accessibility checks builds pages that fail them too. Garbage in, garbage out, now with a monthly subscription.
So the same companies that shipped the inaccessible internet get to sell you the tool that reprints it, then sell you a second tool to fix what the first one broke.
Sources: martech.org · searchengineland.com
Would you like to save this?
We'll email this post to you, so you can come back to it later!