Fascinating exploration. I'd like to see a Turing test of an AI script analysis. Some of the flaws in the ChatGPT coverage occasionally appear in gatekeeper-for-hire services that I've read (people who will cover your script for a fee). That doesn't mean the accountants at a big prodco won't try to cut corners by replacing its readers with a program.

Even if a studio created an AI script reader based on the gurus from the Advice Factory (McKee, Snyder, whoever), I don't think they'd get a much more accurate or detailed review than what they'd get from a decent human reader. In separating the wheat from the chaff, any good human will use intuition based not only on well-written scripts they've read but on movies they've seen (a script may read well but not make a good movie--hence Goldman's cynicism perhaps). They're going to recommend only a script that they think will make a good movie, not just a script that looks good on the computer monitor.

In terms of a writer using AI feedback to improve a script, at best the advice from a machine won't be much better than advice from the Factory, just a little more specific to the script in question. In my experience, assiduously following the advice, templates, etc. recommended by the gurus can (not necessarily will) improve on a mediocre story, but it can't transform a mediocre story into a good story, just one that stinks a little less and does a better job of impersonating a good story. True improvement has to come from the inner workings of a writer who knows the craft and knows how to use feedback. Thousands (probably tens of thousands) of writers have read McKee; we don't have thousands of good screenwriters.