been thinking about that ceiling we always hit with standard audits. most tools can catch missing alt text but they're useless at checking if the text actually makes sense or if focus is trapped in a dialog. saw something about
IGTs trying to bridge that gap by basically doing the heavy lifting and just asking a human for the final call on things like decorative images. supposedly it takes a 45 minute manual check down to maybe two or three minutes if the automation handles the tabbing and logging. sounds great on paper but i'm skeptical about how much
noise this adds to the logs. if the tool is making its own conclusions about what an accessible name should be, we might just end up with a massive backlog of false positives to verify. it's basically shifting the work from finding issues to
verifying automated judgments. hard to see how this scales without a huge amount of oversight on the output quality.
article:
https://www.deque.com/blog/intelligent-automation-and-the-new-era-of-guided-testing/