Is Google Actually Hating on My AI Content, or Am I Just Cooked?

I've developed an AI-powered SEO tool targeting trades businesses such as roofers, HVAC technicians, and plumbers. This tool produces four to five articles daily, utilizing OpenAI's technology. After three months, the results are underwhelming: 168 impressions, three clicks, and an average position of 65.4. While there has been a slight improvement from 73 a few weeks ago, reaching the first page seems like a distant dream.

It's challenging to ascertain whether this is merely a consequence of being a new domain that hasn't earned trust yet or if Google has an issue with AI-generated content, possibly slowing it down covertly.

Firstly, I've addressed the following issues to ensure they're not user errors: duplicate content clusters, which I found and merged manually and with an automated agent. I've also ensured that all articles have proper titles and clear calls-to-action. Furthermore, I've started submitting the content to the Indexing API instead of relying on Google to discover it. Backlinks have been acquired through directories, HARO, and Reddit, albeit at a slow pace.

In my head, I'm grappling with the fact that Google claims to value helpful content, regardless of its creation method. However, the consensus among other builders online suggests that AI content is trapped in "AI jail" until the domain proves its worth. The uncertainty remains whether this claim is true or merely the result of coping mechanisms.

To add to the confusion, my articles are technically accurate and well-structured. Yet, I'm unsure if they read as engaging to a human. This uncertainty might be more critical than the AI versus human debate. Ultimately, I'm wondering if anyone who has successfully shipped AI content at scale has experienced their content being stuck, separate from the usual struggles of a new site that hasn't earned trust.

Or is it always just the domain-age problem? If you generate content quickly without hand-editing every article, what strategies have you found effective to prevent the content from appearing generic without sacrificing speed? I'm trying to determine if I'm addressing the wrong problem at the moment.