There has been a longstanding trend of lowering quality bars.
And most of times, those lowering said bars, don't understand the implications well enough.
AI is giving more incentives to a greater variety of people to lower these bars, in favour of increasing output speed/quantity. And in turn, lowering of the quality bars is helping with proliferation of AI tools in mismatched places.
But the industry seems to have had the "quality" problem long before that.
Mainly thanks to failure of people testing methodologies, which end up being cracked by those, who train primarily for the test (using courses designed to train them only for the given test) and end up being disproportionately (test vs real work) evaluated.
Combine this with failure of workplace testing methodologies and overall company quality testing, we geg what we have now.